Cheating in chess

[Wadih Khoury]

When I correlated ECF and Fide, excluding juniors, the formula is more 7.5x + 600, so a 200 is more of a 2100 Fide. Juniors mess up the correlation a lot.

Regarding Lichess, it is unfortunately not as easy as the population curve is very different than Fide's. At the lower end, it is very routine to find 100/1400 players with a 1800+ rating. In fact, until last year, at my son's level the formula that worked well was lichess = Fide + 400. Now it is probably more of a 200-300 gap (I really doubt he is 2200, and that some of his peers are 2100-2400), and as you get stronger the gap reduces.

Joe, the rule of thumb i use is Ecf (new grade or online grade) = fide +100, lichess = fide +400 and chess.com = fide +200. Not valid at all levels, but gives a decent flavour.

[PeterFinn]

With various different lines of best fit, the relationship has been updated over time, but it has been a constant that 200 equates to 2200. My experience is that the Lichess rating is pretty accurate, possibly 100 points inflated, but not really more than that. Hence, I would be surprised if many 170 players could maintain a 2200 rating.

I had a look at some profiles I know of adults with fairly consistent ECF ratings that had also played a good number of Lichess rapid games. I found:

175 2200
170 2250
200 2350
210 2425

This is obviously not too scientific but gives a rough estimation.

[Pete Heaven]

Wadih,

The WCU has been using 4 digit ratings for decades and, having undertaken a tedious exercise I wished I'd never started a few years ago, I believe our ratings are FIDE+80, give or take, so pretty close to yours. This gives some comfort that we'll be on a levelish playing field for future ECF vs WCU OTB games.

[Joseph Conlon]

Now it is probably more of a 200-300 gap (I really doubt he is 2200, and that some of his peers are 2100-2400), and as you get stronger the gap reduces.

Joe, the rule of thumb i use is Ecf (new grade or online grade) = fide +100, lichess = fide +400 and chess.com = fide +200. Not valid at all levels, but gives a decent flavour.

With various different lines of best fit, the relationship has been updated over time, but it has been a constant that 200 equates to 2200. My experience is that the Lichess rating is pretty accurate, possibly 100 points inflated, but not really more than that. Hence, I would be surprised if many 170 players could maintain a 2200 rating.

My personal data point on this is that I am ~2450 on both rapid and blitz on lichess, and while I would like to believe that I am a much better player than when I last played OTB standard-play chess (with 200 ECF/2200 FIDE), I doubt it. I can believe there is some improvement - my openings are much better than when I was younger, simply because the volume of games I play now is much larger, and I am not having to play myself out of bad positions as I once had to.

I do feel the lichess ratings tighten up quite a lot above 2000 - at the top end, FIDE + 400 clearly doesn't work, as among many GMs there are only two with lichess rapid > 2800. Certainly my experience in either blitz or rapid is that if I play someone 150 points higher than me, I really feel the difference in quality.

[JustinHorton]

I don't know that I believe lichess = fide + 400. Mine is lichess = fide + 158 (though my fide is for normal rates of play and of course that may play a large role) and while I'm sure there's room for improvement - I don't play a lot of rated games, they cause me stress - I don't think I'm equal to players 250 points above me.

If I have a rule of thumb it might replace Wadih's 400 with 200 or so, though at the same time I find that lichess rapid gradings aren't remotely as helpful as estimates of opponents' strength as fide or ecf standardplay ones are. (Well of course they're not, firstly because they fluctuate a great deal and secondly because playing strength presumably fluctuates a great deal more between individual rapid games than individual standardplay games). Also of course a rough calculation that might make sense at lichess 1600 might makes less sense at 2000 or 2400.

[Geoff Chandler]

Hi Roger,

".You could potentially program a chess engine to periodically select a move well down its list of best choices..."

Stealth cheats (nigh impossible to catch) are perfectly capable of doing this without getting a computer to make their choice for them.

A computer does not understand what is a difficult position for a human to play. I cannot see how a computer
can suddenly become human and play a human error. The cunning cheat sees a list of four choices and to vary
picks the 4th if it still said it carried a plus. The trouble with this method ( the good news regarding catching a cheat)
is if they select the 4th choice it may require 10 on the trot 1st choice moves in a messy position to hold the plus.

I like to think of this as a computer setting it's owner a cheats trap - 'go on pick the 4th move and see what happens next.'

[Matthew Turner]

Now it is probably more of a 200-300 gap (I really doubt he is 2200, and that some of his peers are 2100-2400), and as you get stronger the gap reduces.

Joe, the rule of thumb i use is Ecf (new grade or online grade) = fide +100, lichess = fide +400 and chess.com = fide +200. Not valid at all levels, but gives a decent flavour.

With various different lines of best fit, the relationship has been updated over time, but it has been a constant that 200 equates to 2200. My experience is that the Lichess rating is pretty accurate, possibly 100 points inflated, but not really more than that. Hence, I would be surprised if many 170 players could maintain a 2200 rating.

My personal data point on this is that I am ~2450 on both rapid and blitz on lichess, and while I would like to believe that I am a much better player than when I last played OTB standard-play chess (with 200 ECF/2200 FIDE), I doubt it. I can believe there is some improvement - my openings are much better than when I was younger, simply because the volume of games I play now is much larger, and I am not having to play myself out of bad positions as I once had to.

I do feel the lichess ratings tighten up quite a lot above 2000 - at the top end, FIDE + 400 clearly doesn't work, as among many GMs there are only two with lichess rapid > 2800. Certainly my experience in either blitz or rapid is that if I play someone 150 points higher than me, I really feel the difference in quality.

It would appear that I have got this wrong. I have just played a few blitz games and looked up my opponent's rating where possible. In rough terms their Lichess ratings were 100, 300, 200, 200 above their OTB rating.

[JustinHorton]

As it happens, and to some degree contradictory to my comments above, I just played a lichess game against a player with a rating of 2112 whose FIDE rating is advertised in their profile as 1230! (I checked the rating, it's correct as stated. The player's year of birth is 2004.) Their playing strength didn't seem out of line with their lichess rating, though it did with their FIDE rating. They were plainly playing properly.

[Tim Harding]

As it happens, and to some degree contradictory to my comments above, I just played a lichess game against a player with a rating of 2112 whose FIDE rating is advertised in their profile as 1230! (I checked the rating, it's correct as stated. The player's year of birth is 2004.) Their playing strength didn't seem out of line with their lichess rating, though it did with their FIDE rating. They were plainly playing properly.

We need a crucial piece of information here: when did this `16/17 year old last play a FIDE rated game? I have not have played one since January 2020 and many players' FIDE ratings may be even more out of date.

When playing OTB is readily available again, we shall probably see lots of adults lose 30-50 points through being rusty and from playing juniors who have been constantly active online.
We shall certainly see many juniors making huge gains. We shall probably also see juniors who never played FIDE events before shooting up from nowhere to 2000+ especially with their high K coefficient.
Of course it's also possible that some juniors doing well online may not find OTB so easy because they have got in the habit of playing too fast and employing trappy openings that don't hold up so well at slower time controls.

[PeterFinn]

As it happens, and to some degree contradictory to my comments above, I just played a lichess game against a player with a rating of 2112 whose FIDE rating is advertised in their profile as 1230! (I checked the rating, it's correct as stated. The player's year of birth is 2004.) Their playing strength didn't seem out of line with their lichess rating, though it did with their FIDE rating. They were plainly playing properly.

I'm aware of one English junior who has played over 5000 rapid games on Lichess in the last year (possibly one of the players Wadih was referring to when he said he doesn't know how they do it). His published ECF is under 100 and at the beginning of the pandemic his ratings on Lichess were in the 1500s. He has improved consistently month on month and his bullet, blitz and rapid ratings are all now all over 2300. He is maybe the most extreme example of at least a dozen juniors who are approaching 180 strength (at least at rapid chess) but looking at the most recent ECF ratings you wouldn't have a clue.

[John McKenna]

[Okay, it’s not chess, but how often do well trained mid-teen juniors bolt on 400-500 points of performance strength in 10 months? (Not to mention without being able to play any over the board chess.)

I think its also worth noting that improvements are logarithmic in terms of amount of effort - each gap between, say, 2600 to 2400, or 2400 to 2200, or 2200 to 2000, requiring much more effort than the previous one. Going from 1000 to 1600 I can easily imagine....

Yes, for the purposes of this example, player X has bridged the gap from say 1900 to 2400, to the point where they are routinely beating/and or drawing a range of GM's, IM's and FM's. Non-blitz loses are now a rarity.

This was the situation in the 2 decades following Fischer's development (he became a GM at age 15 years in 1958) when meaningful measurements first became possible -

Data on the development of young US masters since 1960 (when the USCF adopted the Elo Rating System)... the group of 12 early achievers includes Browne, Tarjan & Rogoff, who were already titled..." The other 9 were all considered title candidates.

(Prof. A Elo, 1978)

Walter Browne b. 1949, became a GM in 1970
James Tarjan b. 1952, became a GM in 1976
Kenneth Rogoff b. 1953, became a GM in 1978

Av. USCF ratings of group at various ages -

1640 - 12 yrs
1785 - 13 yrs. (145 Elo points increase)
1930 - 14 yrs. (145 Elo points increase)
2120 - 15 yrs. (190 Elo points increase)
2255 - 16 yrs. (135 Elo points increase)
2315 - 17 yrs. ( 60 Elo points increase)
2350 - 18 yrs. ( 35 Elo points increase)
2405 - 19 yrs. ( 55 Elo points increase)
2425 - 20 yrs. ( 20 Elo points increase)
2445 - 21 yrs. ( 20 Elo points increase)
2445 - 22 yrs. ( 0 Elo points increase)

(Prof. A Elo, 1978)

Therefore it would appear that some of the best US junior improvers of 1960s & 1970s could achieve a maximum of about 200 Elo points per year from an average base of about 1640 at the age of around 12 years...

Will the lockdowns result in better or worse (offline) FIDE rating improvements for today's developing talents in the future? Only time will tell.

[Nick Burrows]

There was a junior playing in the London League who is 1550 Fide / 140 ecf. His captain said he believes he is now 2100+.
I checked his Lichess ratings and they had basically gone from 1700 to 2300 at all time controls over 1 year.

[JustinHorton]

We need a crucial piece of information here: when did this `16/17 year old last play a FIDE rated game?

January 2019, I think, and it may be that the rating is based on relatively few games.

I'm aware of one English junior who has played over 5000 rapid games on Lichess in the last year (possibly one of the players Wadih was referring to when he said he doesn't know how they do it).

I've been a member for a year and a week, and I've played 3400...

[Tim Harding]

I've been a member of Lichess for xix and a half months and played 19 games, the majority in two Irish-organised tournaments at 45 minutes with 15 or 30 second increment. I've played somewhat more on chess.com but only because that was the platform for the ECF events at Christmas.

[PeterFinn]

I've been a member for a year and a week, and I've played 3400...

That's a lot, fingers crossed you see a good improvement in your play when otb chess is back then.