Hikaru's Cheating Scandal Got Crazier

Рет қаралды 2,011

Күн бұрын

Пікірлер: 51

@obscurity3027 9 ай бұрын

I’m no mathematician, but I believe that a chess boxing match between Kramnik and Hikaru would draw an enormous audience!

@aesaehttr 9 ай бұрын

If you're not a mathematician then you're a good PR campaign manager. The riches are in niches!

@infomatters. 9 ай бұрын

In the words of the famous Hunger Games actor Charlie aka Moistcr1tical "just because you are good in a board game doesn't mean you are smart" 💀

@jacksarkisian 9 ай бұрын

kramnik put all his skill points into chess 😭

@CloudoCR 9 ай бұрын

Kramnik (skill issue virgin) should just confront Hikaru at this point, some chess boxing could be fun

@jacksarkisian 9 ай бұрын

Kramnik by TKO round 3

@Skunguss 9 ай бұрын

i am hikarus great grandma, hikaru is such a sweet young man and would never do any of this. kramdick is fake news and doesn’t even exist.

@pixzlated 9 ай бұрын

Bro's hair got shaved for the second time in a semester 💔💔💔

@jacksarkisian 9 ай бұрын

no more ellen degeneres comments now 🙌🏻🙌🏻

@pixzlated 9 ай бұрын

dat buzz cut good af tho @@jacksarkisian

@andriusambrutis343 9 ай бұрын

I studied math myself and agree with a Polish guy. In general, I watch Hikaru play from time to time and I find it hilarious. He streams every game, he pre-move a lot and sometimes you cannot even keep track + he always explains his thinking. I just don't see how he would be able to cheat. Maybe Russian player just had bad day?

@Kingzzxepic 9 ай бұрын

Kramnik is unhinged

@Skunguss 9 ай бұрын

is that ellen in the bottom right corner

@scottekoontz 9 ай бұрын

9:49 Using actual Elo win% and a 55.5/56 scenario (all wins, one tie) a Monte Carlo simulation shows only a 21% chance of occurrence within 55,000 games. The issue is that if we trust Elo win% then this not likely. It is also silly to assume that a 55/55 and a 45.5/46 just so happen to ave occurred within months. I do not think Hikaru cheated, but if statisticians are going to lay down some values, then put them all down. Only by using a win% above .85 do you get the values claimed. Elo of +350 above opponents results in 0.79 win, 0.10 loss, 0.20 tie. Yes, the results are sensitive to the win%, but assuming all ties go to Hikaru is just not happening. Of course there are such things as great days, but that is true for opponents. Sure he can pick an opponent, realize they are having a bad day, keep playing them. But what about the next player who is only 150 Elo below him? Etc.

@azeemuddinkhan923 9 ай бұрын

But Hikaru has 5-6 such streaks recently

@azeemuddinkhan923 9 ай бұрын

What is probability of such 5 streaks in games Hikaru played in last 2 months?

@scottekoontz 9 ай бұрын

@@azeemuddinkhan923 Probability is near nil unless Hikaru's blitz Elo is much higher than shown and/or there is a difference in win/loss/draw percentages in blitz.

@LegallyBlue 9 ай бұрын

It's a shame to see Kramnik destroy his well-built legacy over such an absurd set of events, in such absurdly low time. I suppose not everyone deserves a praising footnote in the annals of history...

@Witty_AlienClips 9 ай бұрын

Thank you so so much for making those videos and keeping us updated! Thank you

@danielshapiro2472 9 ай бұрын

The reality of this situation is that this has been a matter of *elementary school level* statistics from the beginning. I remember reading about the coin flip example when I was literally 9 years old at math camp. Essentially, if you have 100 coin flips, you are *overwhelmingly* likely to get at least one streak of 5-6 heads or tails in a row. The likelihood is so high in fact, you can confidently identify a fake set of data if there are not more than 3 in a row during the 100 flips. So yes, as the mathematician states, it would be unlikely for Hikaru NOT to get 46 wins in a row eventually. Let that sink in. Kramnik has less understanding of statistics than an elementary schooler. Not discrediting the mathematician but you do NOT need to be a professor to understand this concept.

@scottekoontz 9 ай бұрын

By the same elementary school math it is *very* certain you get more than a 1 draw in chess after 55 games for the known Elos. *Not* getting a draw within 55 games has a probability of 0.000005% It would be like tossing a die 55 times and never getting a 1. The issue is one of win% in which the difference between 80% and 90% makes a world of difference. The mathematician's work was correct if you use the wrong parameters. In regards to streaks within long sets of trials/games, let's look at the 50/50 with the actual win% for the stated Elos, and not the win% presented by the mathematician because his work ignores draws. For an Elo difference of 350 the win/loss/draw percentages are 79/1/20. If you add half of the draw % to the win% you get to a generic win% used in chess because you get ½ point per draw. The mathematician used 90 for the win% in order to prove it was probable after many trials. Even 55 success in 55 trials with p=.79 => 0.000002% probability of occurrence. But as you said we need to consider a reasonable number of trials. With 1,000 games the odds are 0.0005 or 0.05%. Simply not going to happen. For 10,000 probability is 0.5%. If every single game he ever played was against -350 Elo and in blitz then the probability of a 55/55 is about 10%. In statistics we are looking for 50% to make the even at least as likely as not. If we are talking about the 45.5/46 (all wins and at most one draw in 46) then we get only a 4% probability that such an event happens after 1,000 games, and 36% probability after 10,000 games. Let's totally game the system in Hikaru's favor. After 20,000 blitz games against -350 Elo players we get a slightly better than even chance that he could have possibly achieved a 45.5/46, but no way he would discover a 55/55. So what is happening most be that his blitz Elo is really higher than reported. Don't think he cheated, but Kramnik's math is probably closer to the Elo math than the "mathematician".

@obscurity3027 9 ай бұрын

@@scottekoontz you make a great point. When not accounting for draws, the math is completely different.

@TheFartoholic 9 ай бұрын

@@scottekoontz (1) You're forgetting that we're calculating a probability for an unlikely event BECAUSE it's an unlikely event. If you prospectively toss a die 55 times and never get a 1, that's surprising. If you toss a die thousands of times and find a notable streak within that, you shouldn't be surprised that it was unlikely - that's the reason we're talking about it in the first place. This has to be baked into whatever assessment we make based on probability calculations. (2) Chess games shouldn't be considered independent draws - you have days / weeks where you play better than others making streaks more likely. (3) We have to consider the priors - given how much he streams, how quick he plays - what are the odds that he could actually be cheating? Innocent people have gone to jail for "1-in-a-million" events (google Sally Clark), because they ignore the priors.

@scottekoontz 9 ай бұрын

@@TheFartoholic I'm not forgetting anything because I'm using no assumptions about the known parameters. Statistics can be about surprisingly common outcomes and surprisingly rare or unheard of outcomes after we know the parameters. Hearing people claim 45.5/46 and 55/5 should be common are missing the math and/or statistics. You don't even need stats calculations since Monte Carlo simulations will do. "If you prospectively toss a die 55 times and never get a 1, that's surprising." More than surprising... simply not going to happen. Note that p=0.83 is higher than the odds of winning 55 chess games in a row, because we are talking about zero losses and zero ties which comes to p < 0.80. "If you toss a die thousands of times and find a notable streak within that, you shouldn't be surprised that it was unlikely" In this case we are talking about a specific notable streak: a streak of 55 1s (or any other number). You cannot go looking for any and all notable streaks because that's not the topic. The "notable streak" you are referring to is the probability that there will be a streak of 55 successes (no 1s, i.e. winning 55 chess games in a row). Here is the rub: You should be *very* surprised if you rolled a die 1,000 times and there was such a streak. A 55/55 in chess /may/ happen with enough games and enough players using the known parameters, but just after a 45.5.46? "you have days / weeks where you play better" Streaks are often the helpers for streaks, and there's room for that, but not THAT much room. Hikaru would need not only his better plays *in a row* but also that his opponents are not having a good day *and* no good games above their rating against him. Not saying he cheated. Not at all. Saying a few things: It is possible his blitz Elo is higher than reported, it is possible blitz games have fewer draws that Elo predicts, and the "mathematician" who claim this should be a common occurrence are ignoring the possibility of ties and gaming the win% absurdly towards Hikaru to get to those conclusions.

@TheFartoholic 9 ай бұрын

@@scottekoontz I didn't mean that your calculations were wrong - or that the original calculations were right. Just that there are additional factors that complicate it. When you see low probabilities for unlikely events - you're often calculating probabilities because they're unlikely. That doesn't mean you ignore the probability, but you have to realise that you'd be making the same calculation if it was a streak of 54, 53, ..., etc. So the question shouldn't be "what's the probability of a streak of 55", but "what's the probability of a high streak." The point about correlated draws will depend very much on the facts. It doesn't seem too wild that a person could be playing well above their normal level for a day or two. It happens in all sorts of sports. The 79% figure would apply for a random sample of Hikaru's games, but less so for a consecutive sequence. A better way of doing this would to estimate length of streaks (and its distribution) for a 350 Elo differential, then get the probability of getting higher than 55 out of that distribution (while acknowledging the point raised in the previous paragraph).

@bassmanjr100 9 ай бұрын

Kramnik should go back and finish high school math before discussing high level mathematical statistics.

@fintan9218 9 ай бұрын

I get why hikaru was probably emotional in the moment but why does he think Nepo was agreeing with Kramnik? Seemed more to me he was making fun of Kramnik?

@jacksarkisian 9 ай бұрын

think Hikaru mentioned his relationship with Nepo is pretty poor to begin with & they don’t like eachother. just an educated guess i suppose

@Skunguss 9 ай бұрын

ya know me personally i wouldn’t take that hikaru

@RF-xj1ej 9 ай бұрын

This is such rudimentary probability that anyone who took a probability 101 course in college would be able to immediately see how absurd the claim made by kramnik that is "not against anyone" is.

@jacksarkisian 9 ай бұрын

i think the notion that the run is unlikely is completely valid. its an unlikely run 100%. but even if the chance of it happening was 0.1%, thats still 1/1000. if you play every day on stream like hikaru does (and has major incentives to do now he's on kick) you'd be surprised it doesn't happen more often.

@RF-xj1ej 9 ай бұрын

@@jacksarkisian exactly

@danielshapiro2472 9 ай бұрын

Bro forget about college, I learned that in elementary school math camp no cap 😂

@RF-xj1ej 9 ай бұрын

@@jacksarkisian I just did a simulation where a person has a 88% win rate and play 1000 games. The probability of having a 50 win streak is 18%. Not at all a small chance. This is assuming every game is independent of each other, i.e. the player would be playing different opponents each game. However I believe Gothamchess did an analysis on Hikaru's games and obviouisly Hikaru would play an opponent for like 10 wins straight then another for another streak. I think it is safe to assume in this case the probability of winning will keep going up against the same opponent. So if I assume that the player will have an increased 1% chance of winning, up to 90%, and reset to 88% after losing a game, the probability of having a 50 win streak is significantly higher, to around 40%. Not at all am I suggesting that this is a correct model, just an illustrative example of how this minor assumption can completely change the outcome.

@scottekoontz 9 ай бұрын

@@RF-xj1ej You are correct that 88%, play 1000 games, 50 win streak does in fact return a 18% probability of occurrence. BTW no simulation needed for streaks just something like Wolfram Alpha will do.) It means it would be rare occurrence, but it could happen. Better is to note that it would take 3,500 games to see a 50-50 chance that such a streak would occur. But note that the actual win%, according to Elo of the 45.5/46 streak was between 0.72 and 0.79, not 0.88. 88% has been thrown around and it is wrong since it is common in chess to include half of the draw%. Win rate is 0.79 by assuming Hikaru's Elo was his ending Elo and noting the average Elo difference of 360, which is a fair assumption and generously helps Hikaru. It is 0.72 if we take the averages per each game, where the Elo difference averaged 308. 79%, play 1000 games, 50 win streak would result in a 0.15% probability of happening. 15,000 trials to get to 2% probability (still very unlikely to happen). Kramnik's claims that this is interesting are not absurd. The 45.5/46 was filled up with a 55/55, so no need to use all 35K games Hikaru has ever played.

@Skunguss 9 ай бұрын

idk about this bro

@rainer-lax2967 9 ай бұрын

Im a engineer and scored 77th in a national maths test so maybe half a mathematecian Interesting but no one cares Just the same as hikaru winning against lowel elo player Intereseting but no one cares. If he cheats against magnus, then thats interesting.

@Skunguss 9 ай бұрын

nitro

@CallumOKane 9 ай бұрын

Yet another great video :P

@jacksarkisian 9 ай бұрын

thanku broski ❤️

@scottekoontz 9 ай бұрын

Two kinds of statistics: 1) Elo, Elo difference, win%, simple statistics and probability of the events. 2) Combing through millions of games to see if these are truly rare. 1 is relatively easy, and is what Kramnik is using. And yes, the stats show that 45.5/46 would be extremely rare and would require 18,000 successive games for such a streak to be more likely than not, but the 55/55 is basically a statistical impossibility. So far I have not seen anyone work out the probabilities. I have, but for some reason they are ignored in favor of those "but Hikaru doesn't cheat, I know it in my heat" comments. I don't think he cheated, but I can see where Kramnik is coming from. Mocking the work Kramnik is using is silly because he hasn't even show it yet, but I am coming to the same results so he's not wrong to call 'sus' as Hikaru has in the past. 2 is going to be interesting. Sure, it would be neat to see if such streaks are somewhat common, but we will never know why. Was there cheating, or does the Elo need to be revisited for blitz games?