My explanation as a PhD candidate in Physics: The Glicko system is based on certain mathematical assumptions. These, first of all will only be approximately satisfied with actual human players. For example, a 50 game winning streak against an equally rated opponent may be possible if the opponent is drunk. Secondly, the model‘s predictive power, in terms of predicting probabilities, may only be valid for approximately equally rated opponents, say +- 200 points. If there is a large rating difference, for example 1000 points, the probabilities you calculate may not at all match with the actual probabilities. In a essence, it is just a mathematical model, which is unable to be accurate under all circumstances due to the complexity of human beings. Mathematics is mathematics, but if you apply it to humans it’s power will be limited. Same with mathematical models in physics, all of them have their limitations and you can not expect to always get the right probabilities due to the complexity of the system. Due to these reasons, I would be very careful to draw premature conclusions about cheating.
@@VBKramnik Because the games are not independent events as the model assumes. Hikaru is playing opponents specifically chosen for these kinds of streaks because it's good for stream. Nobody else is playing a lot lower rated players in long matches. If Magnus had an incentive to play for as much streaks as possible, he would absolutely be able to do similar streaks as Hikaru. And you don't have to take my word for it. The author of the mode who you refer to even brings it up at the end. This is from the thread: "As you can see, there are many examples of a player reaching a streak in reality, but the model thinks this streak has a low probability of happening. I believe there are several factors at play here, most importantly the IID assumption. If two players play many games in a row then it's unlikely that these are truly independent events. They may play the same opening many times and get into similar positions. If one of them is better at navigating these specific positions then this may lead to more streakiness than an IID model can explain. There are many other reasons why consecutive games may not be independent. Another shortcoming of my model is the formula used to compute W/D/L probabilities from rating difference. I think a better way to do this is to fit an actual statistical model to real data, to estimate how rating translates into probabilities." It's an interesting statistic, but it doesn't prove anything. Even the author clearly states that this doesn't prove any cheating.
@@VBKramnik Nakamura's playing habits are completely different from any other top player, no other top player spends hours playing online, no other top player plays significantly lower rated opposition for fun, and no other top players cherry picks his opponents as much as Hikaru. (The guys spends hours online, if he only played top players he'd be burned out). He himself is an anomaly.
My hypothesis is that the platform is manipulating pairings of selected players and/or paying players to lose games against selected players. In other words, the "content" produced on the platform is not organic, it follows a partial script with the goal to create "drama" and "incredible performances".
Most people would not believe Hikaru would cheat or collude. Understandable. And I am not saying it. But I do remember how shocked everyone was when Lance Armstrong was caught cheating. Before then everyone thought he was simply the best and attributed his success to rumors (hearsay) such as: Lance was born with a larger heart, etc. But in reality he was simply blood doping.
I watch Hikaru and have often seen him actively choose opponents based off of how "good" they are. For example he can play one game against someone and afterwards he says "he's too good", and then starts searching for some other opponent. I guess he is actively searching for weak high rated players. I don't know if this is something that all the players in the list do, but if not, it could be something that makes Hikaru stand out in statistics like these.
It's not so easy to find a player in the tilt with such frequency. I don't believe that Hikaru's abilities here are very special. But what I do believe is that Hikaru's opponents just intentionally lose to him. Such an anomalous amount of streaks cannot be achieved just by evaluation bar or something like that. It's simply negotiation between players, which is also not fair, of course, but it's another type of cheating after all.
@@omaraboulasse9698 Magnus had 12 10-game unbeaten streaks and a single 25-game unbeaten streak. The probabilities are for this entirely; 78% for 1 25-game streak, 66% for 12 10-game streaks.
at 6:09 when Vladimir is presenting Mike Kagansky you can hear Hikaru saying "he doesn't know what he's talking about'. How did that happen??? what is the chance for it to happen??? editor prank?
Don't a lot of the Hikaru streaks include adoptions and stuff? If you do the math 425 of Hikaru's 2174 (almost a quarter) sample games came from 25 win streaks. This seems to me a clear outlier from farming.
52:00 when a Russian speaker says "Op-puh", you know stuff is going to get juicy (except if he's a chess streamer, in which case he's about to spectacularly blunder)
Владимир хорошо говорит на английском, правда с ОЧЕНЬ РУССКИИ АКЦЕНТОМ, что только радует что Владимир хорошо развит, и что он чисто русский, и мало был в Америке, и не нахватался американского акцента, в общем наш человек
Suppose Hikaru plays 74 independent games. If the probability he wins one game is p, then the probability he wins exactly 72 is (74*73/2*1)*p^72*(1-p)^2. If you let p=31/32, then you get a probability of 0.268 (i.e., 26.8% @2:45). The question is whether you believe p=31/32 is possible. An ELO spread of 600 would allow for it. I'm willing to believe cheating is a problem. If you want to persuade me with mathematics, please make precise claims, not vague insinuations. Suspicions are not evidence.
I'm a passionate Hikaru hater, but the problem with those stats at the end - at least for me - are that as far as i know only Hikaru does the farming of opponents thingy. I would LOVE to get those state for Hikaru with either a) all games removed when he is playing the same opponents more than once in a row or b) all but the first game of a series of games against the same opponent removed. If those number tell the same story i might get started to be convinced that something might be going on there Because the "tilt doesn't matter" argument + those "farm sessions" combination just doesn't do it for me i must say
Yeah, for me its especially the tilt Argument. I am usually pretty consistent with my rating, but it is so easy to fall like 50 points in one session whrn you are angry
I think we have to be careful of extraneous variables affecting some of these calculations. I am a firm believer in elo inflation (artificial inflation of elo), so I feel that could have a major impact on the quality of said players.
I think there's something wierd with the stats on Hikaru. for the 10 games streaks at 52 to be within 93% likely. You only get a chance at a 10 games streak when your last game was a loss. If I assume he lost maybe 10% of games??? that would be 217 oppourtunities and he converts 52 (25% ish) of them to 10+ streaks you say 92% likely. But then looking at the 25 game streaks. You first need to have had a 10 game streak (then win 15 more games). He has had 52 oppurtunities and converts 17 of them (32% ish) and suddenly the likelyhood is down to 0.1%? The likelyhood of winning 10 games after a loss and winning 15 games after a 10 game streak shouldn't be so different. It looks like a mistake in the analysis potentially. Anyone know how many losses in this period? to give a better estimate than my 10%?
@@VBKramnik OK, I've misspoken. But if you replace the word loss with loss or draw, doesn't the same question hold true? Unless his loss/draw is alot over 10%. It's about the coversion rate of 10+ streaks to 25+ streaks, of whatever kind of streak you are measuring. 31% of 10+ win streaks converted to 25+ 40% of 10+ unbeaten streaks converted to 25+ 33% of 10+ overperfrom streaks converted to 25+ It would be interesting to see his conversion rate from a chance to start a streak to getting a 10+ streak. I'm note saying this exonerates him from cheating at all, but it might provide more context in which way. Or if his conversion rate from 0 streak to 10+ streak is in roughly the same area 30-40% then why would there be such a high difference in the likelihood of the 25+ streaks compared to the 10+ Do you have the number of times he drew or lost total from the data series? If you were to estimate would 10% be too high or low?
Mr Kramnik, Are you deleting comments that are critical of your conclusions? A commenter pointed out that a general blitz rating (obtained by playing, eg, 5|3) does not necessarily correspond to performance in the 3|0 time control. This comment seems to have disappeared…
because he doesnt know, or pretends, majority of players just plaiyng 3 0 control there I do not need here comments from those who doesnt even bother checking before writing or misleading on purpose
you and the people liking your comment are just dumb zoomers. world champion vladimir is free to delete any comments he wishes. it's a privilege for us to even comment on his videos and have him respond from time to time
Aren't your stats just showing that the rating model does not work, as potential cheating of Hikaru should be reflected in his rating and therefore the propabilities should never be that low, as the propabilities are calculated with rating differences?
There is an easy way to see if the farming Argument a lot of people are bringing up is valid: If the performance/accuracy of those same players Hikaru is playing doesn’t drop significantly during the streak, the argument is invalid. Tilting or not, they are still high rated players, it won’t turn 0,1% into 50%.
That is a good suggestion for a test. For the probability of a streak, small differences in performance from game to game can turn into quite large differences in the total probability of the streak.
@@sebastianbrandt4556 right,finishing the test, publishing asap. Makes it clear factually that TILTING is completely unjustified concept, numbers are showing the opposite in fact
если Накамура после этого бежит под детектор, а в прошлый раз к нему в гости приходили, то борьба ВБ не просто имеет смысл, но ещё и результат, ну, от этих пройдох, большего и не ожидалось.
Да выучите вы уже этот английский. 500 слов выучите и этого будет достаточно чтобы понимать. Либерасты учите из-за любви к западу. Ватаны учите язык врага 😂 всем можно мотивацию найти :) сорри, не про политику я вообще :) новый ролик под обед самое то.
@@sleepyjoe9386 Проблема не в незнании, просто русскую речь приятней слушать и как-то душевней) а английский, (особенно у ВБ не очень хорошее произношение) режет слух
@@TopUser2022согласен, что приятнее слушать русскую речь. Но завязывайте уже с этим понтами относительно акцента и знания языка. Я 20 лет уже почти в Британии живу. Всем пофиг на язык и на произношение. Главное чтобы понятно было. Я каждый день тут слышу такие акценты что диву даешься. Наши люди только могут цепляться к акценту. С другой стороны интереснее когда ты можешь полноценно следить за всей индустрией. Даже с ботами Накамуры можно посраться :)
I guess it would be useful to find out the player/s he played against to have those streaks. It could be the case the difference in rating is considerable bigger than the one shown in the third column for the overall number of games played.
The GODs of chess (engines) are sophisticated search engines combined with state-of-the-art STATISTICAL models. Yet, we find it hard to rely on statistical analysis of players performance to determine if they are cheating in the game or not 😂
Mr Kramnik, I support your efforts at spotlighting the real problems of online cheating. However, tilt/psychological factors of online chess can't be brushed aside simply because there are a lack of studies on the impact of these factors. Never before could a random 2300 even play 1 game against a top-5 GM. The pressure on said 2300 vs Hikaru would not be pre baked into 2300's elo. It seems clear that the models are incomplete bc the assumptions are incomplete, therefore the probabilities don't carry a ton of weight. Certainly not enough to ban. Surveillance? Prove it again publicly? Sure, maybe. Open to (non-costly) ideas there based on existing technology.
For Hikarus overperformances, the 3 low probalities factors are in reality the same factor., winning streak is an unbeaten streak and an overperformance streak. So in general, from the 25 unbeaten streaks, 21 were overperformance streaks and of those 17 were winning streaks. If the rest of statistical factors don't rise questions. Then the search needs to be narrowed to the 17 winning streaks. Factors to see, as mentioned in the video, long streaks of games against opponents that have lower performances than their ratings, picking opponents to get the streak, analyse the centipawns loss for Hikaru play and opponents play. Another interesting factor that I wish was talked about is the duality between the winning streaks and the unbeatten streaks. And it makes sense, it all comes to drawing percentage. A player that makes more draws, has les probability of winning streaks, but higher probability of unbeaten streaks. a player (like Alireza) that always plays for the win, then will have more winning streaks than its rating, but less unbeaten streaks as well. The rating system makes a big average of that and they even out.
Крамник выиграл у Каспарова и стал ЧМ, вошел историю. Почему Крамника , который защищает шахматы и этому отдал всю свою жизнь, имеет нападки ? Те. тот который своими способностями и огромными знаниями достиг вершины , подвергается нападкам, сомнениями от тех .кто и близко не имеет тех качеств. Это настолько абсурдно, а по сути сводит на нет шахматы и всех тех . кто вписал себя в эту историю.
Rating systems don't have great predictive accuracy...Why are you taking the ratings as reasonable assumptions for your statistical model? ELO is kinda a stochastic process sort of estimator - it's not meant to give an exact probability of winning. If you really wanted to do this you'd need to include the uncertainty in the predictive accuracy of the rating system. It's not impossible - you could set up a STAN program to do it but it's hard. Until you do that these "probabilities" you give are pretty meaningless.
Nothing gives an exact probability of winning, and using ELO is completely fine. If you instantly dismiss all conclusions that were not derived with exactly your preferred method, you will almost never learn anything. Of course what you probably do is even more dangerous: you instantly dismiss anything that makes you feel bad
Numbers are good, but where is the actual method of getting them? The model shows some good results, but it certainly does not handle the extreme cases. Also would be interesting to see what happens if the same sample size is equal for everyone, for example 400 games.
Just one remark, if a player scores a streak of let say 20 games and the probability is let say 50% it can't be sad that he scores like every second game in that isolated exempel in term of prosent. If there is a sample of 1000 games, 20 games are only 2%. If there is a several number of straeaks can the equation predict the next outcome and the probability for future streak ? Does the probability apply for the next 1000 games? How does this statistic correspond to the chance in term that outcome of games can be like a result in turning the dice?
@@vlaer81 Your comment reads like a Hikaru fan: poor emotional control resulting in nonsensical outbursts. Do remember sarcasm doesn’t come through in text.
@@vlaer81 Reviewing the time stamp I realize now you were referring to the Hikaru cut in, which I didn’t realize was there. My mistake, and I apologize. I’m leaving my comments up for the record to clarify the situation if anyone reads this thread.
Thanks champ! Halfway through now and interested to see the statistics! I taught a statistics class back in the day ;-) Please never give up on ignoring the haters and fighting cheating! You are awesome!!
I have just one explanation, except the obvious one. All those muppets who lose Hikaru in those streaks, they lose on purpose to earn his respect and the respect of that famous cartel. This is how he farms this rating, this is how he collects these streaks and this is how he promotes cartel. Muppets got that cartel's respect and everyone is happy. Bots on the frontline who are rated around 600 won't even look at this stat. Numbers are complicated for them. I spotted a few of them already in comments.
The farming is the easiest explanation, of course. And as for me numbers are speaking very well here. Anomalous probabilities are only in long streaks 25+ games. It's obvious farming.
I can confirm when hikaru broke the blitz record I witnessed multiple games where in the end the winning underdog gives up a queen or allowed hikaru to win on time when 1 move draws to simply capture hikaru s last pawn were declined and hikaru wins on time.
@@thegodpill9696 I saw a lot of games when opponents are losing on purpose it seems to him. Including money events. Is world of chess already so corrupted?
@@VBKramnik even if we are not talking about actual cheating with computer engine, still this sort of manipulation with rating should be punished, but seems like no one cares, except us. They need a hero with all these streaks. Nothing personal, just a marketing and business.
The problem is that players who cheat usually let themselves win 2 or 3 games between their victories to hide it, they even make relatively bad moves to lower their accuracy.
Писдец, Владимир говорит на английском, и еще ебать субтитры на английском, я и так не понимал, и в ДВОЙНЕ ничего не понимаю. Владимир, делайте пожалуйста видео на русском, я их точно буду смотреть, возможно я один, но я точно буду.
I have recently searched about the effectiveness of statistics and in fact many scientists and mathematicians encourage everyone to learn about it. Its a shame the chess world doesnt take it seriously
Владимир Борисович, держитесь! Обидно, что гроссмейстеры бояться открыто поддержать ваше исследование и так робко выражают обеспокоенность ситуацией. Ведут себя как настоящие "хата с краю"
The statistics seem to indicate not cheating, but an intentional blending of ratings. Hikaru's statistics are anomalous only for streaks of 25 games. But if Nakamura's streaks always involve different people, there is no other way to explain it besides cheating. This is quite convincing, as is the method of checking by simulation.
Only if you can trust the model used. Basic probability is pretty worthless if you ignore all of the other potential variables like nerves, tilt, etc. etc. These numbers don't mean anything and using a flawed model doesn't give it credibility in the absence of a better model
I don't think that basic probability is worthless and such factors as tilt, bad mood, inspiration make the big difference on the long distance. 2000+ games are enough to make conclusion about obvious anomaly.
@@thefunk8398ok, so everything is worthless, we don't have to waste time anymore. Numbers mean nothing, because they don't consider all details. It's clear. Thank you.
Is it just me or did someone edit Hikaru talking over in this video, for example here 16:22. There are multiple interruptions like this throughout the video. This is not really funny.
So I just finished watching the video and the interruptions are almost at every 5 minutes. I don't even know what to say, it's just sad that people will go to such lengths to try and minimise the impact of others' opinions.
@@kuney6518 there are just a few episodes where you can hear Hikaru and even then the voice is very easy to miss. What are you talking about? Are you mental?
@@vlaer81 check the time spams posted in comments, someone clearly edited Hikaru voices and clips over this video to belittle what Kramnik was trying to say.
Hikaru with the 1 in 1000 amount of streaks... Absolute nonsense, except when you play with a headset on, huh? Can Maurits the mathematician do it again for Hikaru, but for the last full year instead?
But to actually add something meaningful to this discussion, I would refer to this mathematician's own admission, and correct assumption from an academic open-minded point of view, that his model could be flawed. In research, the investigator's job is not to back up their hypothesis with what data they can find, this mindset can lead to unintentional/unconscious cherry picking of data/numbers. Instead, they should be their own harshest critic, so that they can create a robust research design and recognize the possible limitations of their design. I'm not sure if we can agree on this, but it would be a good point to find common ground.
Now I don't know what extraneous variables you and your team may or may not have considered, so I'd like to offer a few. Some you may have already briefly considered, but not intuitively agree that they affect the study. I will try to convey why I think they should be seriously considered. A little precursor, these variables matter because while numbers are a very useful way to understand the world in a seemingly perfect way, they are only as good as the variables we can account for, as the world and it's systems are rarely as perfect as the numbers we use to model them. This is why weather forecasts are seldom completely accurate. I argue that the online chess ecosystem is one such system.
I want to present these variables in three sections, that I think capture them pretty well. 1. Farming (and why rating may not be the full story) Online blitz ratings are much more volatile than fide ratings, even fide blitz, as such many players experience highs and lows of 150-250 points or more. You might argue that this itself is troubling, and we can discuss further if so, but I see it as perfectly reasonable. Choosing the right opponents to farm matters, and this is not necessarily reflected in their objective rating. While we can in theory say that all 2900's should all be objectively equal in strength, I think this over generalisation fails to capture the fact that 2900's can have their own strengths and weaknesses. Some have really strong/solid openings, some may thrive in positional/tactical positions, and some may love to play endgames. Some may be exceptionally good/focused on looking for so called cheap traps and tricks, and some may be really good at flagging their opponents in completely drawn positions. Many may have terrible nerves and get really nervous only when playing renowned players. Of course most are a varying combination of all these things. 'I argue that two players of equal rating are not going to play as well in any given position/situation, in other words, they are not always actually of equal strength'. This is important because Hikaru, playing online for as long as he has, knows which players of a certain rating are "easier" to beat than others of the "same" strength, and knows how to exploit their weaknesses. This can even change for the same player on a different day as evidenced by the fluctuations in rating that many, if not most, online players experience. I have seen Hikaru avoid certain players when farming, not because he thinks they are cheating, but because he thinks they are a bad pairing for him to trick/squeeze/flag. Sometimes a player he chooses to farm one day will play better another day, and he simply chooses not to play them at that moment. He often plays players when they are near their "peak" rating (if he has determined that they are vulnerable, ie not playing their best), to take as many points as possible, but a player's peak rating is not indicative of their "true"/average rating/strength. TLDR: /Ratings don't necessarily capture playing styles that can be exploited. (See Hikaru Nakamura vs Magnus Carlsen lifetime Classical score 1-14 Carlsen) /Ratings don't necessarily present the true playing strength of an opponent who fluctuates 250 points. (See any online player, check their peak and their average rating) /Ratings don't necessarily reflect the mental fortitude of a player under pressure. (See the performance of top players in the Candidates, so I hear) 2. Positions and time I think we can easily agree that bullet results are much more volatile than blitz, and blitz more so than rapid, and rapid more so than Classical. The point is, results become more volatile as time decreases, as in Classical where less time becomes more like blitz and sillier mistakes can be made. This is the ideal condition for the best (one of?) 1+0 online bullet chess player to capitalize on his opponents weaknesses. It's no coincidence that Hikaru exclusively plays 3+0 blitz casually online. As time dwindles down, I would argue that his superior bullet skill over that of his opponents become a bigger factor so that these much more volatile positions can often go his way. I argue this difference in strength under time pressure is not already reflected in the opponent's blitz rating, as players handle their time differently throughout the entire game against similarly rated opposition. In these streaks and over performances, we see only what the numbers say, that he won and against what player, but not the context of the story. Not the many positions that he barely scraped a win by in completely lost positions due to his opponents time pressure. As an example, we can see in sports sometimes, that we say the results don't tell the whole story. In soccer, a team could have won 3-0, and from the result we could conclude that one team was much better, but sometimes one team had many opportunities, played well and put pressure, but the final execution simply didn't go their way. MVL and Nakamura played in one such match. In an SCC, MVL came close many times to beating Nakamura, but in the end some of those games simply didn't go his way in time scrambles, and the end result showed a lopsided dominating result for Nakamura. TLDR: /Simply looking at the results of the games may not be the best way to model the probabilities of volatile results where one player uses the chaos to their advantage. 3. Psychology I already mentioned that some players crack under pressure, be it the pressure of the moment or the pressure on the clock, and this is not necessarily reflected in rating. But this is online blitz, where games come one after another and mood swings can significantly alter the level of play of a player. One bad game, depending on the person, can ruin their mood and significantly lower their ability. This happens to Hikaru, and he sometimes simply decides to take a break, but lower rated players will continue to play while tilted, because they like the opportunity of playing a top player. This tilting can compound, and while a player may be rated 2900+, they could easily play at a level of
Там в начале объясняется метод расчёта вероятностей. 1000 раз повторялась серия. Допустим в каждой из этих 1000 серий были 10 победных стриков из 10 партий. Значит вероятность этого события 100% А вот в симуляции серий Накамуры стрики из 25 побед практически не встречаются, в отличие от реальности. Либо с реальностью что-то не так, либо с симуляцией ) Лично мне метод проверки кажется достаточно убедительным и понятным. Нет заумных формул - просто бросаем монетку 1000 раз и смотрим, какой процент будет у орла, а какой у решки )
With the cheating you are 1000% right!! Pls stop with the other more absurd claims like with the "eval bar" where you are clearly wrong and just undermining your original point on cheating. Good luck!
@@McLarenMercedes17 it's not completely relevant, but just go to the daily puzzle section on the cartel's website. People with 200 rating in rapid they say EZ literally under every puzzle, even when the puzzle is quite complex. Something tells me these people will easily cheat in a real game. There is a ridiculous amount of cheaters in every segment
But there are no Kramnik statistics in this video. This video shows the statistics of a professional mathematician who responded on Twitter to a suggestion to do statistics about Hikaru. Are you calling him pathetic?
Владимир Борисович, спасибо. Вчера встретился канал @Tornado-iz4rd, где человек, как я понял, программист и математик, тоже использовал интересные метрики и уже подтвердил часть Ваших результатов.
Более того он обнаружил такое интересное явление как теневой (скрытый) бан (shodow ban). Это когда чесс-ком не банит аккаунт, но запрещает играть на какой-то период времени, например, на полгода или на год. А затем этот аккаунт может опять играть на чесс-ком. В итоге публично никто толком и не замечает этого явления. Насколько я понял, автор канала обнаружил теневой бан у одного из героев Вашего недавнего видео - у индийского сотрудника чесс-ком, тренера, против которого Ян с трудом удержал ничью.
![Звуковые иллюзии, которые работают на всех (почти) [Veritasium]](http://i.ytimg.com/vi/8pCuUfbdheE/mqdefault.jpg)
My explanation as a PhD candidate in Physics:
The Glicko system is based on certain mathematical assumptions. These, first of all will only be approximately satisfied with actual human players. For example, a 50 game winning streak against an equally rated opponent may be possible if the opponent is drunk.
Secondly, the model‘s predictive power, in terms of predicting probabilities, may only be valid for approximately equally rated opponents, say +- 200 points. If there is a large rating difference, for example 1000 points, the probabilities you calculate may not at all match with the actual probabilities.
In a essence, it is just a mathematical model, which is unable to be accurate under all circumstances due to the complexity of human beings. Mathematics is mathematics, but if you apply it to humans it’s power will be limited. Same with mathematical models in physics, all of them have their limitations and you can not expect to always get the right probabilities due to the complexity of the system.
Due to these reasons, I would be very careful to draw premature conclusions about cheating.
Just curious about the reason it works with everyone else but Nakamura 🙂
@@VBKramnik Because the games are not independent events as the model assumes. Hikaru is playing opponents specifically chosen for these kinds of streaks because it's good for stream. Nobody else is playing a lot lower rated players in long matches. If Magnus had an incentive to play for as much streaks as possible, he would absolutely be able to do similar streaks as Hikaru.
And you don't have to take my word for it. The author of the mode who you refer to even brings it up at the end.
This is from the thread:
"As you can see, there are many examples of a player reaching a streak in reality, but the model thinks this streak has a low probability of happening. I believe there are several factors at play here, most importantly the IID assumption. If two players play many games in a row then it's unlikely that these are truly independent events. They may play the same opening many times and get into similar positions. If one of them is better at navigating these specific positions then this may lead to more streakiness than an IID model can explain. There are many other reasons why consecutive games may not be independent. Another shortcoming of my model is the formula used to compute W/D/L probabilities from rating difference. I think a better way to do this is to fit an actual statistical model to real data, to estimate how rating translates into probabilities."
It's an interesting statistic, but it doesn't prove anything. Even the author clearly states that this doesn't prove any cheating.
@ojoj-c6r there are untill the opposite is mathemathically proved. So do it if you disagree. Period
@@VBKramnik Nakamura's playing habits are completely different from any other top player, no other top player spends hours playing online, no other top player plays significantly lower rated opposition for fun, and no other top players cherry picks his opponents as much as Hikaru. (The guys spends hours online, if he only played top players he'd be burned out). He himself is an anomaly.
@@andresv.8880PROVE it mathematically otherwise it is just an opinion, and NOTHING else
I think Hikaru is simply just choosing which opponent to play a lot of the time, you can see it in his stream.
Полностью поддерживаю Владимира Крамника
My hypothesis is that the platform is manipulating pairings of selected players and/or paying players to lose games against selected players.
In other words, the "content" produced on the platform is not organic, it follows a partial script with the goal to create "drama" and "incredible performances".
Как же похорошел стендап при Крамнике
Сам ты 🤡
I do believe that a lot of online chess players cheat. I'm with you on this Mr Kramnik.
Most people would not believe Hikaru would cheat or collude. Understandable. And I am not saying it. But I do remember how shocked everyone was when Lance Armstrong was caught cheating. Before then everyone thought he was simply the best and attributed his success to rumors (hearsay) such as: Lance was born with a larger heart, etc. But in reality he was simply blood doping.
Or motor doping!
You either die a hero, or live long enough to become the villain
I watch Hikaru and have often seen him actively choose opponents based off of how "good" they are. For example he can play one game against someone and afterwards he says "he's too good", and then starts searching for some other opponent. I guess he is actively searching for weak high rated players.
I don't know if this is something that all the players in the list do, but if not, it could be something that makes Hikaru stand out in statistics like these.
It's not so easy to find a player in the tilt with such frequency. I don't believe that Hikaru's abilities here are very special. But what I do believe is that Hikaru's opponents just intentionally lose to him. Such an anomalous amount of streaks cannot be achieved just by evaluation bar or something like that. It's simply negotiation between players, which is also not fair, of course, but it's another type of cheating after all.
@@staspivovartsev5579 it is even worse, total corruption, if so
Most don't because most GMs don't care about their online rating. Hikaru seems to care a great deal about his.
@@stoutlager6325 there's much easier ways to keep the online rating high
@@omaraboulasse9698 Magnus had 12 10-game unbeaten streaks and a single 25-game unbeaten streak. The probabilities are for this entirely; 78% for 1 25-game streak, 66% for 12 10-game streaks.
at 6:09 when Vladimir is presenting Mike Kagansky you can hear Hikaru saying "he doesn't know what he's talking about'.
How did that happen??? what is the chance for it to happen???
editor prank?
miracle
Don't a lot of the Hikaru streaks include adoptions and stuff? If you do the math 425 of Hikaru's 2174 (almost a quarter) sample games came from 25 win streaks. This seems to me a clear outlier from farming.
52:00 when a Russian speaker says "Op-puh", you know stuff is going to get juicy
(except if he's a chess streamer, in which case he's about to spectacularly blunder)
Владимир хорошо говорит на английском, правда с ОЧЕНЬ РУССКИИ АКЦЕНТОМ, что только радует что Владимир хорошо развит, и что он чисто русский, и мало был в Америке, и не нахватался американского акцента, в общем наш человек
Suppose Hikaru plays 74 independent games. If the probability he wins one game is p, then the probability he wins exactly 72 is (74*73/2*1)*p^72*(1-p)^2. If you let p=31/32, then you get a probability of 0.268 (i.e., 26.8% @2:45). The question is whether you believe p=31/32 is possible. An ELO spread of 600 would allow for it.
I'm willing to believe cheating is a problem. If you want to persuade me with mathematics, please make precise claims, not vague insinuations. Suspicions are not evidence.
I'm a passionate Hikaru hater, but the problem with those stats at the end - at least for me - are that as far as i know only Hikaru does the farming of opponents thingy. I would LOVE to get those state for Hikaru with either a) all games removed when he is playing the same opponents more than once in a row or b) all but the first game of a series of games against the same opponent removed. If those number tell the same story i might get started to be convinced that something might be going on there
Because the "tilt doesn't matter" argument + those "farm sessions" combination just doesn't do it for me i must say
Yeah, for me its especially the tilt Argument. I am usually pretty consistent with my rating, but it is so easy to fall like 50 points in one session whrn you are angry
You should then also remove farming for all of the other players to make it a fair comparison
@@zonk45 that's nothing, I've been bouncing around 1650+ to 1900 for the past year lol
I think we have to be careful of extraneous variables affecting some of these calculations. I am a firm believer in elo inflation (artificial inflation of elo), so I feel that could have a major impact on the quality of said players.
Владимир Борисович, когда четвёртая часть Пешек и Королей? Или вы хотите снять вторую трилогию, как Питер Джексон?
I think there's something wierd with the stats on Hikaru. for the 10 games streaks at 52 to be within 93% likely. You only get a chance at a 10 games streak when your last game was a loss. If I assume he lost maybe 10% of games??? that would be 217 oppourtunities and he converts 52 (25% ish) of them to 10+ streaks you say 92% likely. But then looking at the 25 game streaks. You first need to have had a 10 game streak (then win 15 more games). He has had 52 oppurtunities and converts 17 of them (32% ish) and suddenly the likelyhood is down to 0.1%?
The likelyhood of winning 10 games after a loss and winning 15 games after a 10 game streak shouldn't be so different. It looks like a mistake in the analysis potentially.
Anyone know how many losses in this period? to give a better estimate than my 10%?
Draw wouldnt fit? Think before writing, but first ask you school teacher, will explain you
@@VBKramnik OK, I've misspoken. But if you replace the word loss with loss or draw, doesn't the same question hold true? Unless his loss/draw is alot over 10%.
It's about the coversion rate of 10+ streaks to 25+ streaks, of whatever kind of streak you are measuring.
31% of 10+ win streaks converted to 25+
40% of 10+ unbeaten streaks converted to 25+
33% of 10+ overperfrom streaks converted to 25+
It would be interesting to see his conversion rate from a chance to start a streak to getting a 10+ streak.
I'm note saying this exonerates him from cheating at all, but it might provide more context in which way. Or if his conversion rate from 0 streak to 10+ streak is in roughly the same area 30-40% then why would there be such a high difference in the likelihood of the 25+ streaks compared to the 10+
Do you have the number of times he drew or lost total from the data series? If you were to estimate would 10% be too high or low?
Mr Kramnik, Are you deleting comments that are critical of your conclusions?
A commenter pointed out that a general blitz rating (obtained by playing, eg, 5|3) does not necessarily correspond to performance in the 3|0 time control. This comment seems to have disappeared…
because he doesnt know, or pretends, majority of players just plaiyng 3 0 control there
I do not need here comments from those who doesnt even bother checking before writing or misleading on purpose
you and the people liking your comment are just dumb zoomers. world champion vladimir is free to delete any comments he wishes. it's a privilege for us to even comment on his videos and have him respond from time to time
Aren't your stats just showing that the rating model does not work, as potential cheating of Hikaru should be reflected in his rating and therefore the propabilities should never be that low, as the propabilities are calculated with rating differences?
No those are probabilities according to his rating, that is the sad (4u) reality
There is an easy way to see if the farming Argument a lot of people are bringing up is valid: If the performance/accuracy of those same players Hikaru is playing doesn’t drop significantly during the streak, the argument is invalid. Tilting or not, they are still high rated players, it won’t turn 0,1% into 50%.
That is a good suggestion for a test.
For the probability of a streak, small differences in performance from game to game can turn into quite large differences in the total probability of the streak.
@@sebastianbrandt4556 right,finishing the test, publishing asap. Makes it clear factually that TILTING is completely unjustified concept, numbers are showing the opposite in fact
если Накамура после этого бежит под детектор, а в прошлый раз к нему в гости приходили, то борьба ВБ не просто имеет смысл, но ещё и результат, ну, от этих пройдох, большего и не ожидалось.
Владимир Борисович, делайте побольше видео на русском!
Да выучите вы уже этот английский. 500 слов выучите и этого будет достаточно чтобы понимать. Либерасты учите из-за любви к западу. Ватаны учите язык врага 😂 всем можно мотивацию найти :) сорри, не про политику я вообще :) новый ролик под обед самое то.
Поставьте яндекс браузер с озвучкой перевода видео
@@sleepyjoe9386
Проблема не в незнании, просто русскую речь приятней слушать и как-то душевней) а английский, (особенно у ВБ не очень хорошее произношение) режет слух
@@TopUser2022согласен, что приятнее слушать русскую речь. Но завязывайте уже с этим понтами относительно акцента и знания языка. Я 20 лет уже почти в Британии живу. Всем пофиг на язык и на произношение. Главное чтобы понятно было. Я каждый день тут слышу такие акценты что диву даешься. Наши люди только могут цепляться к акценту. С другой стороны интереснее когда ты можешь полноценно следить за всей индустрией. Даже с ботами Накамуры можно посраться :)
@@sleepyjoe9386
Никаких понтов. В Британии всем, может, и пофиг, а я говорю за себя. Также и русский какого-нибудь таджика неприятно слушать.
I guess it would be useful to find out the player/s he played against to have those streaks. It could be the case the difference in rating is considerable bigger than the one shown in the third column for the overall number of games played.
I have it all, sometimes SLIGHTLY bigger, but sometimes smaller, like with other players there
Kramnik is extremely intelligent. Wonder if he will study maths and statistics at University? He will be brilliant just like in chess.
The GODs of chess (engines) are sophisticated search engines combined with state-of-the-art STATISTICAL models. Yet, we find it hard to rely on statistical analysis of players performance to determine if they are cheating in the game or not 😂
Mr Kramnik, I support your efforts at spotlighting the real problems of online cheating.
However, tilt/psychological factors of online chess can't be brushed aside simply because there are a lack of studies on the impact of these factors. Never before could a random 2300 even play 1 game against a top-5 GM. The pressure on said 2300 vs Hikaru would not be pre baked into 2300's elo. It seems clear that the models are incomplete bc the assumptions are incomplete, therefore the probabilities don't carry a ton of weight.
Certainly not enough to ban. Surveillance? Prove it again publicly? Sure, maybe. Open to (non-costly) ideas there based on existing technology.
For Hikarus overperformances, the 3 low probalities factors are in reality the same factor., winning streak is an unbeaten streak and an overperformance streak. So in general, from the 25 unbeaten streaks, 21 were overperformance streaks and of those 17 were winning streaks. If the rest of statistical factors don't rise questions. Then the search needs to be narrowed to the 17 winning streaks. Factors to see, as mentioned in the video, long streaks of games against opponents that have lower performances than their ratings, picking opponents to get the streak, analyse the centipawns loss for Hikaru play and opponents play.
Another interesting factor that I wish was talked about is the duality between the winning streaks and the unbeatten streaks. And it makes sense, it all comes to drawing percentage. A player that makes more draws, has les probability of winning streaks, but higher probability of unbeaten streaks. a player (like Alireza) that always plays for the win, then will have more winning streaks than its rating, but less unbeaten streaks as well. The rating system makes a big average of that and they even out.
39:50 This laugh 😂
I request Mr. Kramnik on behalf of chess community to play world blitz and analyse the difference between online and offline chess better.
Very interesting and convincing video. Thank you very much, Mr. Kramnik.
Крамник выиграл у Каспарова и стал ЧМ, вошел историю. Почему Крамника , который защищает шахматы и этому отдал всю свою жизнь, имеет нападки ? Те. тот который своими способностями и огромными знаниями достиг вершины , подвергается нападкам, сомнениями от тех .кто и близко не имеет тех качеств. Это настолько абсурдно, а по сути сводит на нет шахматы и всех тех . кто вписал себя в эту историю.
Rating systems don't have great predictive accuracy...Why are you taking the ratings as reasonable assumptions for your statistical model? ELO is kinda a stochastic process sort of estimator - it's not meant to give an exact probability of winning. If you really wanted to do this you'd need to include the uncertainty in the predictive accuracy of the rating system. It's not impossible - you could set up a STAN program to do it but it's hard. Until you do that these "probabilities" you give are pretty meaningless.
Nothing gives an exact probability of winning, and using ELO is completely fine. If you instantly dismiss all conclusions that were not derived with exactly your preferred method, you will almost never learn anything. Of course what you probably do is even more dangerous: you instantly dismiss anything that makes you feel bad
Do it then
Interesting
Numbers are good, but where is the actual method of getting them? The model shows some good results, but it certainly does not handle the extreme cases. Also would be interesting to see what happens if the same sample size is equal for everyone, for example 400 games.
Just one remark, if a player scores a streak of let say 20 games and the probability is let say 50% it can't be sad that he scores like every second game in that isolated exempel in term of prosent. If there is a sample of 1000 games, 20 games are only 2%. If there is a several number of straeaks can the equation predict the next outcome and the probability for future streak ?
Does the probability apply for the next 1000 games?
How does this statistic correspond to the chance in term that outcome of games can be like a result in turning the dice?
This was such a nice informative video. Thanks Kramnik
What I also notice is how Hikaru always has his headset on his head
52:06 🤣🤣 literally nothing here 🤣
Typical Hikaru fan
@@trumppence3834 explain your comment and don't forget to apologise for this terrible insult
@@vlaer81 Your comment reads like a Hikaru fan: poor emotional control resulting in nonsensical outbursts. Do remember sarcasm doesn’t come through in text.
@@vlaer81 Reviewing the time stamp I realize now you were referring to the Hikaru cut in, which I didn’t realize was there. My mistake, and I apologize. I’m leaving my comments up for the record to clarify the situation if anyone reads this thread.
@@trumppence3834 don't worry :) np
Thanks champ! Halfway through now and interested to see the statistics! I taught a statistics class back in the day ;-)
Please never give up on ignoring the haters and fighting cheating! You are awesome!!
Bullet Brawl is a good tournament to catch them all
I have just one explanation, except the obvious one. All those muppets who lose Hikaru in those streaks, they lose on purpose to earn his respect and the respect of that famous cartel. This is how he farms this rating, this is how he collects these streaks and this is how he promotes cartel. Muppets got that cartel's respect and everyone is happy. Bots on the frontline who are rated around 600 won't even look at this stat. Numbers are complicated for them. I spotted a few of them already in comments.
The farming is the easiest explanation, of course. And as for me numbers are speaking very well here.
Anomalous probabilities are only in long streaks 25+ games. It's obvious farming.
@@staspivovartsev5579or cheating
I can confirm when hikaru broke the blitz record I witnessed multiple games where in the end the winning underdog gives up a queen or allowed hikaru to win on time when 1 move draws to simply capture hikaru s last pawn were declined and hikaru wins on time.
@@thegodpill9696 I saw a lot of games when opponents are losing on purpose it seems to him. Including money events. Is world of chess already so corrupted?
@@VBKramnik even if we are not talking about actual cheating with computer engine, still this sort of manipulation with rating should be punished, but seems like no one cares, except us. They need a hero with all these streaks. Nothing personal, just a marketing and business.
The problem is that players who cheat usually let themselves win 2 or 3 games between their victories to hide it, they even make relatively bad moves to lower their accuracy.
Реклейминг термина grandpa и вставки реакций Хикару - хорошая стратегия 👍
Владимир Борисович , а ролики про ваши любимы партии будут в этом году еще или вы полностью сконцентрированы на эту тему ?
Писдец, Владимир говорит на английском, и еще ебать субтитры на английском, я и так не понимал, и в ДВОЙНЕ ничего не понимаю. Владимир, делайте пожалуйста видео на русском, я их точно буду смотреть, возможно я один, но я точно буду.
😂
Good stats, good content, good edit
I have recently searched about the effectiveness of statistics and in fact many scientists and mathematicians encourage everyone to learn about it. Its a shame the chess world doesnt take it seriously
Very interesting video keep up with those videos
Владимир Борисович, держитесь! Обидно, что гроссмейстеры бояться открыто поддержать ваше исследование и так робко выражают обеспокоенность ситуацией. Ведут себя как настоящие "хата с краю"
так они все на западе считай живут, а тут как бы даже мужик может сказать, что он женщина и все буду молча обеспокоены :) такая вот тут жизнь :)
How could I miss this awesome video 🤣
Good job! Потренировал своё понимание английского.
The statistics seem to indicate not cheating, but an intentional blending of ratings. Hikaru's statistics are anomalous only for streaks of 25 games. But if Nakamura's streaks always involve different people, there is no other way to explain it besides cheating. This is quite convincing, as is the method of checking by simulation.
Only if you can trust the model used. Basic probability is pretty worthless if you ignore all of the other potential variables like nerves, tilt, etc. etc. These numbers don't mean anything and using a flawed model doesn't give it credibility in the absence of a better model
I don't think that basic probability is worthless and such factors as tilt, bad mood, inspiration make the big difference on the long distance. 2000+ games are enough to make conclusion about obvious anomaly.
@@staspivovartsev5579 A flawed model is worthless. If you don't factor in all the variables, then the result is wrong. The numbers don't mean anything
@@thefunk8398ok, so everything is worthless, we don't have to waste time anymore. Numbers mean nothing, because they don't consider all details. It's clear. Thank you.
@@thefunk8398 Newton physics is worthless for macro, because it doesn’t include quantum mechanics :?) You so funny in your believes (not knowledge)….
Is it just me or did someone edit Hikaru talking over in this video, for example here 16:22. There are multiple interruptions like this throughout the video. This is not really funny.
Just watched this moment. This is really funny . Go cry to you daddy nakamura :)
@@sleepyjoe9386 another moment here at 22:32 , there were few more before this. Idk how they got to edit this, it's weird how this went unnoticed.
So I just finished watching the video and the interruptions are almost at every 5 minutes. I don't even know what to say, it's just sad that people will go to such lengths to try and minimise the impact of others' opinions.
@@kuney6518 there are just a few episodes where you can hear Hikaru and even then the voice is very easy to miss. What are you talking about? Are you mental?
@@vlaer81 check the time spams posted in comments, someone clearly edited Hikaru voices and clips over this video to belittle what Kramnik was trying to say.
I am enjoying watching your math Journey. You are a super GM in chess. Your math coach is the equivalent of a math IM.
Vladimir, I saw on reddit, that you participated in late TT, how about stream next time?
|ˈmjuːtʃʊəlɪ| мьючюэли - the better way of pronouncing this
Ваще по барабану. Все все поняли. Видел кучу теоретиков которые потом попадали в британский паб и не могли три слова связать в нем.
Thanks a lot for the videos in english :)
Hikaru with the 1 in 1000 amount of streaks... Absolute nonsense, except when you play with a headset on, huh? Can Maurits the mathematician do it again for Hikaru, but for the last full year instead?
What if theymeant by 0.01℅ is too low it should be like 70℅ instead and he is underperfoforming is unlikely
Gotcha checkmate to all pets ()😂😂😂😂
This is undefendable position for pets
Well done sir 👍
I am a critic of your work (not the intention but the implementation), but the Hikaru edits were pretty funny.
But to actually add something meaningful to this discussion, I would refer to this mathematician's own admission, and correct assumption from an academic open-minded point of view, that his model could be flawed. In research, the investigator's job is not to back up their hypothesis with what data they can find, this mindset can lead to unintentional/unconscious cherry picking of data/numbers. Instead, they should be their own harshest critic, so that they can create a robust research design and recognize the possible limitations of their design. I'm not sure if we can agree on this, but it would be a good point to find common ground.
Now I don't know what extraneous variables you and your team may or may not have considered, so I'd like to offer a few. Some you may have already briefly considered, but not intuitively agree that they affect the study. I will try to convey why I think they should be seriously considered.
A little precursor, these variables matter because while numbers are a very useful way to understand the world in a seemingly perfect way, they are only as good as the variables we can account for, as the world and it's systems are rarely as perfect as the numbers we use to model them. This is why weather forecasts are seldom completely accurate. I argue that the online chess ecosystem is one such system.
@@andresv.8880 so please do it and offer your calculations I will check
I want to present these variables in three sections, that I think capture them pretty well.
1. Farming (and why rating may not be the full story)
Online blitz ratings are much more volatile than fide ratings, even fide blitz, as such many players experience highs and lows of 150-250 points or more. You might argue that this itself is troubling, and we can discuss further if so, but I see it as perfectly reasonable.
Choosing the right opponents to farm matters, and this is not necessarily reflected in their objective rating. While we can in theory say that all 2900's should all be objectively equal in strength, I think this over generalisation fails to capture the fact that 2900's can have their own strengths and weaknesses. Some have really strong/solid openings, some may thrive in positional/tactical positions, and some may love to play endgames. Some may be exceptionally good/focused on looking for so called cheap traps and tricks, and some may be really good at flagging their opponents in completely drawn positions. Many may have terrible nerves and get really nervous only when playing renowned players. Of course most are a varying combination of all these things. 'I argue that two players of equal rating are not going to play as well in any given position/situation, in other words, they are not always actually of equal strength'. This is important because Hikaru, playing online for as long as he has, knows which players of a certain rating are "easier" to beat than others of the "same" strength, and knows how to exploit their weaknesses. This can even change for the same player on a different day as evidenced by the fluctuations in rating that many, if not most, online players experience. I have seen Hikaru avoid certain players when farming, not because he thinks they are cheating, but because he thinks they are a bad pairing for him to trick/squeeze/flag. Sometimes a player he chooses to farm one day will play better another day, and he simply chooses not to play them at that moment. He often plays players when they are near their "peak" rating (if he has determined that they are vulnerable, ie not playing their best), to take as many points as possible, but a player's peak rating is not indicative of their "true"/average rating/strength.
TLDR:
/Ratings don't necessarily capture playing styles that can be exploited. (See Hikaru Nakamura vs Magnus Carlsen lifetime Classical score 1-14 Carlsen)
/Ratings don't necessarily present the true playing strength of an opponent who fluctuates 250 points. (See any online player, check their peak and their average rating)
/Ratings don't necessarily reflect the mental fortitude of a player under pressure. (See the performance of top players in the Candidates, so I hear)
2. Positions and time
I think we can easily agree that bullet results are much more volatile than blitz, and blitz more so than rapid, and rapid more so than Classical. The point is, results become more volatile as time decreases, as in Classical where less time becomes more like blitz and sillier mistakes can be made. This is the ideal condition for the best (one of?) 1+0 online bullet chess player to capitalize on his opponents weaknesses. It's no coincidence that Hikaru exclusively plays 3+0 blitz casually online. As time dwindles down, I would argue that his superior bullet skill over that of his opponents become a bigger factor so that these much more volatile positions can often go his way. I argue this difference in strength under time pressure is not already reflected in the opponent's blitz rating, as players handle their time differently throughout the entire game against similarly rated opposition. In these streaks and over performances, we see only what the numbers say, that he won and against what player, but not the context of the story. Not the many positions that he barely scraped a win by in completely lost positions due to his opponents time pressure. As an example, we can see in sports sometimes, that we say the results don't tell the whole story. In soccer, a team could have won 3-0, and from the result we could conclude that one team was much better, but sometimes one team had many opportunities, played well and put pressure, but the final execution simply didn't go their way. MVL and Nakamura played in one such match. In an SCC, MVL came close many times to beating Nakamura, but in the end some of those games simply didn't go his way in time scrambles, and the end result showed a lopsided dominating result for Nakamura.
TLDR:
/Simply looking at the results of the games may not be the best way to model the probabilities of volatile results where one player uses the chaos to their advantage.
3. Psychology
I already mentioned that some players crack under pressure, be it the pressure of the moment or the pressure on the clock, and this is not necessarily reflected in rating. But this is online blitz, where games come one after another and mood swings can significantly alter the level of play of a player. One bad game, depending on the person, can ruin their mood and significantly lower their ability. This happens to Hikaru, and he sometimes simply decides to take a break, but lower rated players will continue to play while tilted, because they like the opportunity of playing a top player. This tilting can compound, and while a player may be rated 2900+, they could easily play at a level of
@@andresv.8880 Not really, dont have enough free time writing the reasons making EACH point of yours not valid in fact, but thank you for sharing
Total vindication for Vladimir Kramnik. [By the way, love your final table where we can easily see everything in one place.] Hikaru is dead to me.
Interesting
Vladimir, for any scientific educated person your arguments are demolishing. Keep relying on statistics in front of trash talking. All the best!!!!!
39:53 😂
great work!
That Maurits thread? OMG. Hikaru's stats stand out like a very sore, very red, thumb.
WORLD CHESS ♟️ CHAMPION 🏆💪 V KRAMNIK WE SUPPORT YOUR WORK & INVESTIGATIONS 1000% BRAVO 👏👍 MY FRIEND!!!! 🙏📿✝️💯🎯
KRAMNIK WITH THE HOT TAKE ONCE AGAIN!!
great work....keep it up
Let's gooo
что за 100 процентные шансы на 40 минуте?
Там в начале объясняется метод расчёта вероятностей. 1000 раз повторялась серия. Допустим в каждой из этих 1000 серий были 10 победных стриков из 10 партий. Значит вероятность этого события 100%
А вот в симуляции серий Накамуры стрики из 25 побед практически не встречаются, в отличие от реальности. Либо с реальностью что-то не так, либо с симуляцией ) Лично мне метод проверки кажется достаточно убедительным и понятным. Нет заумных формул - просто бросаем монетку 1000 раз и смотрим, какой процент будет у орла, а какой у решки )
Колдун.жпг))
With the cheating you are 1000% right!! Pls stop with the other more absurd claims like with the "eval bar" where you are clearly wrong and just undermining your original point on cheating. Good luck!
11/10
Это вы наложили комменты Хикару поверх видео?)
sami nalosilis
@@asdkpi6897 скорее всего, это тот же модератор, который наложил градусник на стримы Хикару )
😯
Владимир Борисович, мы с вами!!
Ну что, Хикару. Лицо к осмотру. Причина тряски?
Владимир Борисович, поддерживаю, но жалко Вашего времени и нервов. Берегите себя
Pathetic.
you sure are
@@Momus2024 I sometimes feel like he is creating a lot of accounts to support himself, because noone does
@@Kaorii1905 you're just like Kramnik who believes many people are cheating
@@McLarenMercedes17 it's not completely relevant, but just go to the daily puzzle section on the cartel's website. People with 200 rating in rapid they say EZ literally under every puzzle, even when the puzzle is quite complex. Something tells me these people will easily cheat in a real game. There is a ridiculous amount of cheaters in every segment
But there are no Kramnik statistics in this video. This video shows the statistics of a professional mathematician who responded on Twitter to a suggestion to do statistics about Hikaru. Are you calling him pathetic?
Владимир Борисович, спасибо. Вчера встретился канал @Tornado-iz4rd, где человек, как я понял, программист и математик, тоже использовал интересные метрики и уже подтвердил часть Ваших результатов.
Более того он обнаружил такое интересное явление как теневой (скрытый) бан (shodow ban). Это когда чесс-ком не банит аккаунт, но запрещает играть на какой-то период времени, например, на полгода или на год. А затем этот аккаунт может опять играть на чесс-ком. В итоге публично никто толком и не замечает этого явления. Насколько я понял, автор канала обнаружил теневой бан у одного из героев Вашего недавнего видео - у индийского сотрудника чесс-ком, тренера, против которого Ян с трудом удержал ничью.
@@vitalykiselev3951 да,благодарю. Посмотрю
теневой бан, конечно знаю