Our first experience with online cheating was traumatic to say the least.
In the early Covid lockdowns we setup an online chess tournament, not expecting much and were blown away when we had to close entries two weeks later after 300 players had paid their entry fee!
It was an online junior event in age groups from Under 8 to Under 18.
This was as low-stakes an event as you can possibly imagine.
We saw no incentive for anyone to cheat. No prizes to be won, no rating points to be gained, the teams concept was new so there was very little prestige as well. A bunch of the players knew each other too which I believed would help.
There was also an apparent huge downside. We were using real names, so players' real reputations were on the line, coaches would find out, team-mates would know. Why risk the embarrasment? Why take a chance at devastating your reputation in a tiny community?
We naiively assumed that with high-risk and low-reward, using real names and live arbiters, there would be no cheating.
We couldn't have been more wrong!
After 5 rounds, accusations started flying. Coaches stress levels rose and I buried my head in Stockfish and PGN Spy and spent hundreds of hours analysing games. It all of a sudden went from the lowest priority for Tornelo to the highest priority. It was a genuine problem for tournament oganisers and needed attention!
Not only did I dive into the game analysis, I also spent dozens of hours on the phone with parents trying to explain what we were doing, listening to their reactions and seeing how players justified their actions, or defended their innocence.
The weirdest thing was that some of the strongest players were getting assistance! In some cases a player 1000 points higher rated still used an engine to play almost every move! Why?
We couldn't understand.
Of course that was May - it's now August and 3 months later we are much wiser.
Firstly, a huge thanks to Professor Kenneth Regan, the absolute world expert at detecting assistance in chess games. I can't even imagine how much work he has done supporting the entire chess world through these last 6 months.
Ken Regan provided us with a Statistical Analysis of the event, of course it didn't say explicitly who cheated, but it gave us a lot of information to work with. Without this report I wouldn't have been able to make any sensible decisions.
Ultimately (2 weeks after the event ended) I made the decision to remove 10 players' results.
The players removed were:
Shortly after I suspected that I'd made a mistake. One of those 10 players was incorrectly removed. I looked too closely at individual game statistics and didn't trust the "big picture" statistics enough. The power of large numbers!
Just last week Tornelo released a Fair Play assessment tool and I thought I'd go back and test my decision making from 2 months ago against the new tool.
This is the report generated by Tornelo (available to arbiters in real time, round-by-round).
All players who were removed were anonymised with a repeated random letter. Players who were not removed appear unaltered.
I'm very excited to see that the Top 9 players on this Tornelo report were all players whom I had chosen to remove. As expected the 10th is way down the list, an unfortunate confirmation that I got it wrong.
This provides reassurance that my decisions at the time were good, but also gives me confidence in the Tornelo Fair Play report. Instead of spending hundreds of hours pouring over game analysis and agonising over a decision, I could have used this in real-time during the event!!
Next time I'll be able to flag suspicious behaviour while the event is in progress stop it before it starts, saving everyone time and heartache!
I'd like to unpack the cheating-in-chess-events phenomonen which is sweeping the globe even more quickly than Covid-19. And offer a solution.
What is cheating and how do players do it?
Cheating is the intentional breaking of rules in order to obtain an advantage over other teams or players.
The very first rule in the FIDE Laws of Chess states:
1.1 The game of chess is played between two opponents
Which is perhaps why this is such an emotional issue - players are breaking the very first Law of Chess!
FIDE Laws get more explicit with:
11.3.1 During play the players are forbidden to use any notes, sources of information or advice
And get into the procedural details in their Anti-Cheating Regulations.
The only form of cheating we are talking about here is players getting assistance from a chess engine.
Why do players cheat?
When defending players who have been accused, parents and coaches will often ask "but, what's the motive?", failing to understand that a motive is not required.
The lure of seeking assistance is incredibly strong and psychologically difficult to resist. We see assistance most frequently with players in their teens and early-mid 20's, unsurprisingly, the phase of life before impulse control has fully developed in the brain. Technically speaking these players are psychosocially immature.
Psychosocial immaturity makes them more likely to:
It becomes even harder to resist in emotional situations, which is why many players start seeking assistance once they get into a bad position, or immediately after a loss.
Getting assistance is the path of least resistance. It's easy and provides huge short-term emotional gains. It's actually harder NOT to cheat! Getting assistance is like eating fatty, salty, sugary foods... both easy and delicious, with short-term rewards and a long-term, almost invisible, downside.
Everyone knows they should eat more vegetables, exercise regularly and play chess without getting any assistance. But very few people consistently do what they know should be done.
We need to be asking how we can help people to NOT get assistance, by creating environments where cheating is easier to resist. Putting players in a high-speed (ie. emotional), anonymous envrionment and expecting them not to cheat is setting them up for failure. It's really the organisers who are being unfair by having unrealistic expectations.
What environmental factors can help?
Authenticity and empathy are the key factors which build trust. A supportive envrionment must ensure all players can be authentic and ideally know and even see one another:
Do you need to catch EVERYONE that cheats?
Most certainly not. We are trying to build trust and create a supportive environment, not to catch cheaters. Game theory has explored this in detail and your community can quickly evolve into a trusting and trustworthy (ie. no cheating) place just by:
What about automated cheat-detection?
By definition getting assistance is only a problem if it leads to an unfair advantage. This means we need an expected performance baseline. If you have no prior knowledge about a player you cannot say if they are cheating or not.
Automated cheat-detection can easily capture Flagrant Cheating (ie. copy every move from the computer), but it can't easily identify Intermittent Cheating (ie. start cheating once you're losing).
Automated cheat-detection causes long-term damage to the community of Trust that you are trying to build. It's anonymous, unforgiving and unempathetic, which may well produce more cheaters than it removes.
How do Communities solve this problem?
Only within a Community will a player get Authenticity, Repeat Interactions and Empathy.
You must be using Real Names and as much as possible be in a community where players know each other.
If cheat-detection isn't automatic, how do we find cheaters?
Arbiters need information in order to make an informed decision. They need information about a player which can help determine an expected performance and they need information about the games played to understand the statistics of what happened in games. That's because the same results from an IM would be acceptable, but from a 650 rated player would be clearly cheating.
A community leader (arbtier) will know the player and know if they are 600 points underrated. They will have a feel for how inconsistent that player is. They will understand a player's mood and the tournament situation and the other players and how they will react and they will be able to have private conversations with parents or players and 'nip it in the bud'.
There will be many times when instantly banning a player is the worst move, not the best move.
The manual identification of cheaters allows:
Tornelo provides a round-by-round process that an arbiter can quickly eliminate those who are not cheating. The arbiter can provide confidence to the higher rated player that, despite it being hard to believe, they just played badly!
You can never build trust in a hands-off Mega-portal tournament full of anonymous usernames, no matter how many cheaters you catch.
But you can quickly build trust in a community of known players with an arbiter that has timely information and control over the process.
Tornelo today released a Fair Play report feature for Arbiters. This will provide round-by-round analysis of games to help Arbiters build trust in their community.
We try not to call it an anti-cheating feature, or claim that we will stop cheating, because everything we are doing is trying to BUILD TRUST between players. It's not about catching the cheaters, it's about making sure everyone trusts one another and that the environment and community is a place where 'it's just not done'.
We don't even want to call it cheating, better to talk about it in terms of "Unfair Assistance"; trust me, parents respond much better when you tell them their precious darling is "Receiving Unfair Assistance during a game", than when you call them a cheater!
We still have a long way to go, so we'd love some feedback on this early feature release!
OK, how does it work?
It all starts by clicking the innocuous "Fair Play" button. Click the button and you'll see a Report.
First we display some overall statistics about how many games have been played in your event, how many have been analyised and how many are still pending analysis.
Tornelo provides 4 levels of game analysis. All games are automatically analysed (Level 1) as soon as they finish. We require you to manually request a deeper analysis.
Tornelo has some threshholds, which if reached, will recommend you to dig a little deeper into the games. The Deeper Analyisis row will describe how many players are being recommended for further investigation.
Simply click the Deeper Analysis button and this will trigger those players to be analysed further.
There are 4 levels of analysis, each with more stringent thresholds. As such, you may see the Deeper Analysis button up to 3 times. You'll need to click it each time if you wish to dig deeper into these players' games.
Below the summary information is a data table. Here is the explanation of each field.
I know now you want the answer to "So, which players are cheating?" ... but for the answer to that you'll have to read another blog post, another time.
Sorry, I’ve always wanted to write a clickbait headline and this was my chance.
It’s impossible to 100% eliminate cheating. But read on and I’ll tell you what to do instead of attempting the impossible.
There are two types of cheating in chess:
It’s Intermittent Cheating that is impossible to prevent and difficult to catch.
At Tornelo.com our objective is not to "prevent cheating". The attempt itself may even encourage more cheaters by giving them a challenge! And a feeling of success when they "get away with it".
Our goal is simple:
Three simple things achieve this:
We all know people behave much worse online than in real life. Behavioural psychology suggests this is due to:
To combat these risk factors Tornelo.com creates an environment in which:
The platform ensures that:
Tornelo.com is a platform for Arbiter Led events. The Organiser builds a community. They know the players. The players know each other. Reputations are built offline and extend into Tornelo.com from the real world.
Arbiters are present and interact with players throughout every event. The arbiters should be explicit about their expectations at the start of every event. Explaining what cheating is, why it’s not ok and creating high expectations - people will usually live up to your expectations (be they good or bad).
Mega-Portals are completely different from Tornelo.com and have a very different challenge. They are a single community of millions of anonymous players. Games are played on-demand, unsupervised and anonymously. They must rely on big-data and statistical analysis to identify cheating.
We don’t need to prevent cheating, we just attract players who wouldn’t cheat in real life and engage them in a community with a culture of fair play and integrity!
We haven’t eliminated cheating, it’s just not the done thing on Tornelo.com.