Playing with anonymous handles is a major problem. It reminds me of the “good” old days on Usenet leading to litigation with She who shall not be named.
According to Nigel Short, even Bobby Fischer played as an anon on ICC.
You keep getting it wrong! Didn’t you read the books? It’s “must” not be named, not “shall.”
– Hal Terrie
I agree many or most top players use handles but the name behind the handle is usually pretty well known and they value reputation and rating, especially if they make money from streaming, or use their high online ratings to attract students. They probably have masked accounts when they are trying out repertoire ideas.
Ask me if I care…
I feel like this issue is going to come to a head sooner rather than later. The USCF has been encouraging its affiliates to take their clubs online. The three top venues seem to be Chess.com, ICC and Lichess, at least based on the webinars that USCF has offered. All of these sites have proprietary approaches to determine cheating which are not transparent to the end user. We (or at least I) don’t really know if they have been examined by the USCF.
But the message from USCF is muddled at best. They have offered USCF rating sanctioned play on two of these sites, ICC for the last seven years or so, and Chess.com for the last year and a half or two years. Both sites seem to have a uniform policy with regard to USCF rated play: if their system discovers a player cheating, that player is banned from future USCF rated play on that site. If I as an affiliate run tournaments there, i have no choice as to whether that player can continue to play there. As the venue account is tied to the USCF ID, it is hard for the player to return under an alias.
It seems too me that the fact that the USCF allows rated play on sites with this policy in effect establishes a defacto policy on the part of the USCF that this practice (the banning of players based on algorithmic analysis of their play) is acceptable. If I, as an affiliate, or a TD, run rated tournaments on one of these sites, what is USCFs guidance to me?
There are two aspects to this. One may be (and i use the term “may be” intentionally, as we don’t know how these systems work) that the gameplay is examined after the tournament is completed. If cheating is discovered, what then? I believe ICC’s policy is to submit the results as recorded, and ban the player. In that case, the results, prize fund if there is one and rating implications stand. Hardly a desirable result if cheating actually occurred. I believe that Chess.com does not submit results to USCF, they leave that thorny problem to the affiliate TD.
The other issue is that improper play may not have occurred continuously or consistently throughout the tournament in question. In my case in the previous paragraph I was thinking of the player that cheated in every game they played in the tournament. Here, suppose we have a 5 round tournament and the system determined that a give player cheated in one game, the game that they played in round two. Had this happened in a OTB tournament (i.e. a player was discovered without a doubt cheating in round two), the player would have been ejected from the tournament. But in the online case, that player played in further rounds. Had they been ejected, the pairings would have changed and the results of the tournament might well have changed as a result. This is another difficult issue that this sort of approach (post game analysis to determine cheating) poses.
What then is the solution? Thus far, it seems as though the USCF has implemented a straightforward but simplistic solution only. The rating of online play is firewalled from a USCF member’s OTB rating. Those two categories are completely broken out from each other. I think that in some ways this works. My experience is that most players take OTB rating more seriously that online ratings. They seem accept that online play is more frangible. Mouse-slips, internet connection failures etc., things beyond your control (including cheating) affect your online rating in ways that just don’t occur (or at least are very infrequent) in OTB play.
I’ve seen many passionate arguments here that get back to one’s life philosophy. They are cogent, and for the most part well argued in a civil way, which is a rarity on the internet and is good to see. I think that chess played remotely across the internet is fundamentally different that OTB, and as such, the set of compromises and rules required to implement it in as a fair a means as possible are likely to be different. I have accepted the fact that a computer can play better chess than me. I may be willing to accept that a computer may be able to detect whether or not I am playing like a human, a reverse Turing test if you will. On that, I remain undecided. But I do recognize that it may be a condition of playing USCF rated chess on the internet should I choose to do so. The bottom line has to do with trust. You either trust that the cheat examining software is accurate or you don’t.
Were my club to offer rated USCF internet play (in the current environment), I think we’d have a responsibility to inform you ahead of time that:
1 - All your games will be examined by the venue’s proprietary software for fair play
2 - A determination by that software that you cheated would be reported to the USCF, and would result in a ban on further play on that platform in accordance with that venue’s guidelines
3 - Games in which it was discovered that you were cheating would be recorded as forfeited, and player standings would be adjusted accordingly
4 - The prizes would only be awarded after the games had been examined, and rankings had been adjusted if need be
You would then have a clear understanding of how play and awarding of prizes would be examined and conducted (much as you do now in OTB play) and could make as informed a decision as possible as to whether you wanted to participate.
To me, this is the current state of affairs, until the USCF weighs in with more explicit guidance, if they ever do. Given that they have been allowing rated play on the internet for seven years and do not have much in the way of explicit and clear guidance, I suspect that it will be some time before they do. Perhaps the pandemic will spur them on, as OTB play is off the table and their revenues dry up. Time will tell.
Very well reasoned. I hope you post here more often. We need more rational thinkers.
Not to disrupt your reasoning, but US Chess has been rating online events on chess.com since March of 2015, a lot longer than two years.
Thanks for the correction, I laid down the facts as best I knew them.
If you play an on-line game on chess.com or any other site and that game is US Chess rated, then if their algorithm says that you cheated and you are kicked off the site you have the option to appeal to US Chess. The appeal would force chess.com (or whoever) to verify their contention that you cheated. In turn, that would force them to divulge how their algorithm works, which so far they have not been willing to do. Upshot - you win the appeal. It’s also worth noting that in those cases of which I am aware when threatened with a lawsuit over this issue chess.com has backed down and reinstated the player. Maintaining confidentiality over how their algorithm functions seems to be their overriding concern.
Every opening has “theory” that has developed over a long period of time and tested in actual play as well as by comparison with chess engines’ assessment of “best play.” If a player plays this known theory for a long series of moves without veering from the main path, how can the algorithms used to check for cheating distinguish between a human player and a computer which can use its opening book? For example, there is a large body of theory in the 6. Bg5 variation of the Najdorf that extends past the 30th move. There are lines in the Exchange Variation of the Grunfeld Defense that go past move 40. These lines have been checked and rechecked by opening theorists and practitioners using computer programs. The same has been true for the Ruy Lopez and French Defense, Winawer Variation. A player with a good memory of these lines can appear to be playing like a computer even if he leaves his silicon monster in another room.
In addition, some endgame theory has been well developed to the point that a well skilled player can rattle off a long series of “best” moves that a computer can find, too. Would a computer call Akiba Rubinstein a “cheater” when examining his rook endgames? Programs have analyzed the famous Lasker-Lasker endgame of N vs R and found that Emanuel Lasker played the “best” move time after time, according to Mark Dvoretsky. Would Lasker be called a “cheater” for finding the same moves with his own brilliant head that the computer found to be extremely accurate?
Programs differ in their evaluations of positions. Their move selection procedures are not perfect, sometimes flipping evaluations once their depth increases through the use of faster processors. What depth do the present algorithms of these playing sites use? What is their horizon? How do they know that the “best” move selected by an algorithm is really the best one? What happens when their are multiple choices that are equal, or a matter of taste and style? How does the algorithm deal with move order transpositions that lead to the same position? Is a player “cheating” by using these transpositions to fool his opponent, even a computer program?
Modern programs since 2010 appear to have passed the “Turing Test”, playing well enough to play like a human. What test is there to show that humans are playing like a computer, to distinguish the play of man and machine? Both man and the silicon monsters play many technical and “book” positions the same way, move for move. Can you really tell the difference from one move, a short series of moves, or even a long series to determine “cheating” when the ideas or plans are well known?
Tom’s posts in this thread (whether intentionally or inadvertantly) explain why:
- ICCF permits engine usage, and all serious correspondence play should, and
- online live play (and correspondence play) is and must be rated separately.
His parable about his student shows a remarkable lack of self awareness. But at the end of the day, it’s online chess, and must be treated in accordance with its limitations and its importance.
Consider: if we had self-awareness, would we be posting here?
Ask me whether I care.
I do play correspondence exclusively on ICCF, and use both Leela and Stockfish extensively. The danger is in ceding control to the beasts: my attitude is that it’s better to go -2 and use one’s brain on occasion than to go +2 & merely take dictation.
(Well, that’s my ATTITUDE: I believe I have yet to go -2 in an event, so I am still taking too much dictation. 
)
Comparison against the results of multiple computer engines would be easy to implement. That being said, I suspect that the algorithms and processes used look at a variety of factors, of which the sequence of moves as compared to what a computer would do is but one. There is a significant amount of ability to track what actions you take with your computer when you use a browser, and there are opportunities for pertinent data collection there. If you use a proprietary program as can be the case with ICC, or with multiple services if a a mobile platform is used, then the opportunity for data collection increases. The analyses of a larger data set, including multiple games played, would also be of value.
All of this, however, is speculation. Which, to me, gets to the core message of your post. I feel like you are unable to get to a point of being able to trust that the detection system works well, because you don’t understand how it works. This is the frustration we all have when there is a lack of transparency.
Suppose the online services were to offer the USCF (which would likely mean a specific cadre of individuals at the USCF which does not yet currently exist at least I don’t think) full transparency, i.e. the ability to visit their facility, get fully briefed, and see the anti-cheating process in action. Suppose, that the USCF then concluded that the process was fair and accurate, and declared such publicly. Two enormous and highly unlikely suppositions, in my view, but follow me down this path for a second. Would you then be willing to play USCF rated games on those sites? My suspicion is that, were this to occur, the nay-sayers would still say no, and the yea-sayers would still say yes; perhaps a few folks on the fence would be swayed in one direction or the other. All this would be is a transfer of trust from the providers to the USCF. Not a transfer of trust to the end user.
Really this is a risk/gain scenario based on trust. If you trust that the systems developed correctly detect cheating then you will risk your online rating and your reputation to play rated games on those sites. If you do not believe it is possible to create such a system, or that the current systems have not yet reached a level of accurately determining such behavior, then you won’t take that risk. This is a frought decision, because you have very little factual data to go by.
That, for me, is the best characterization of the current problem that we all face. There are, of course, other factors that may influence one’s choices. I am an older person with underlying medical issues, significantly more at risk for playing OTB. This may be my only avenue for playing USCF rated chess for an indefinite amount of time. That might influence my decision. We each come to the table with our own level of knowledge, our own set of values, and our own experience. Based on those factors, we make the best decision that we can for ourselves. We may try to influence the people who have power to affect change (that, after all, is really what this forum is about, isn’t it?), but in the end, you have to make the best decision you can for yourself based on the knowledge, experience and values that you have. True for any situation, really.
Which implies that the algorithm is exposed to defeat, if known.
At one time correspondence play banned the use of computers/programs. That was at a time when the programs were not that good and it was easier for a skilled correspondence player to outthink a silicon monster. With the advent of the modern programs, these organizations recognized that many of their players were not going to play on their “honor”. They could not police it or figure out who was using technical tools to play. They threw up their hands and redefined cheating, allowing anything goes. Today, if you use one or multiple programs to help you, it is okay. In effect there is no cheating because it is ignored by the organizations. Since they cannot control cheating they close their eyes and say it does not exist. What a great solution! They claim that quality of play is the more important thing, that correspondence players are expanding the science of chess. Any means justifies the end. Fine. Everyone knows the ground rules and compete in that framework.
In correspondence play people use programs, databases, opening books, latest magazines, correspondence journals, TWIC, Informants, NIC YB, their own little heads, as well as other people to help them. I find it funny that those who think it is okay to use these tools, but then turn around and think it is unethical in a meaningless online game to have someone looking over your shoulder offering general advice in a quick time control. Online play is prone to technical glitches, mouse slips, and an uneven playing field because of the quality of the hardware each player has which affect online ratings. There is an underlying assumption that the rating system is measuring what we want it to measure, that it is pristine, perfect. There is also an assumption, one that is hidden from view, that it is possible to catch computer “cheaters”. Those assumptions are questionable. We have been told that online play is for fun, not to worry about rating points. Now we see that online play is being used to damage reputations without much recourse because there is little transparency. It may eventually come out that the algorithms are subjective, not objective. Did US Cjess make a good deal, exposing its members to specious accusations by a third party organization?
If computer programs are to be banned from use in online play, then it is necessary to clearly define what is cheating and how it is going to be found. If you are going to criminalize play, then define the assumptions that you are using to identify cheaters. Show what policing tools you are using and the rules you are following. Establish a due process system with the same ideal as in our real courts that one is innocent until proven guilty beyond a shadow of a doubt. No accusation of cheating should be made by another organization without US Chess approval after a thorough investigation and defense by the accused. If you cannot and will not do that, hiding behind settlements and NDAs to protect your poor algorithm, then it is better to drop the notion that you can catch cheaters and go the correspondence route where everything goes and then pretend you are pioneering chess as a collaborative scientific effort. The players under these anything goes rules then will play at their own risk and cannot cry they are being cheated. It is ridiculous that players are being accused of cheating in casual online games, and tournaments with no prizes.
Personally, I do not think many players care who or what they play on the other end. They just want to play and do not want to be accused of cheating when they play well. When my former student played, with my minor assistance, we were not the ones who made an accusation, formal or informal, that computers were involved in the play. It was the disgruntled opponent who thought, at the very least, that the dad was helping out because he set up the connection. Our attitude was “So what if he is using a computer? Let’s beat’em anyway.” As it turned out, two humans beat human+silicon monster. That game was a long time ago, well before OTB play disappeared because of the virus. My former student is likely still womping computer assisted opponents, even without my help, as we are not just doing social distancing, but county distancing. It is a shame that the National HS and National Elementary Championships were cancelled. The little tiger was ready to rock and roll.
I suspect there is also a business interest in keeping proprietary work in which the company has perhaps invested significant resources closely held from the competition.
However, to your point, the processes they use are likely to be a scaling effect; as one reaches greater effectiveness with ones detection tool, the adversary often adjusts their methods to defeat that. Looked at through that lens, it is ill-advised to expose to the adversary exactly what you are doing to detect them.
@ tmagchesspgh
I believe the USCF still holds correspondence games to the original standard of no outside help. How they police that is no clearer that the topic at hand. You are correct that the ICCF allows the use of computer assistance.
I respectfully disagree with you here. I believe the the USCF firewalled the ratings for online games for just the reasons you state. Each player in the USCF has three sets of ratings, correspondence, OTB (referred to as “regular”) and online. Rated games one plays in any of these USCF arenas only affect the ratings within that arena, i.e. your online rated games don’t affect your OTB or correspondence ratings, and so forth. It is my belief that this was done because online games were more frangible, just as you state.
I thought USCF was more mainstream. Quoting the 7th edition, chapter 9 p. 293
This is only my interpretation but I would think that consultation of databases (especially precompiled before game start) would likewise be legal - what difference between a precompiled database of openings and MCO 15, except that the database is likely newer?
Lichess, for one, also has a ‘no engines’ rule.
As to how to police that, the same way online sources do. Post-analysis of a game and when one is out of book a human player (even aided by book) can only play so many moves by what a computer defines perfection as. I only play at sites that have a ‘no computers’ rule, but I don’t worry myself much about trying to catch an opponent paired to me… then again, I’ve always been a low rated player.
Ah! By the title, I thought this was a proposed “How To”! 
