Lichess4545 Ledger 117

Lichess4545 Ledger

Issue #117 - October 25, 2019

State of the 4545 League

by @kostasvl

The 4545 league has 285 approved registrations and will will likely see around 30 teams of 8 or 40 teams of 6 or 24 teams of 10 for next season! We will know for sure on Monday.

Side-leagues News

by @kostasvl

Registration for Season 16 of the LoneWolf 30+30 league is open! We're an 11-round 30+30 swiss tournament running from November 5th to January 28th (with a week off for Christmas). We have two sections: U1800 and Open. Our league attracts many of the strongest Classical players on Lichess including dozens of 2000+ and several titled players. It's a great opportunity to get longer-time control games without worrying about your team's results, and it's a great way to play strong opponents you wouldn't otherwise find in the 45+45 league. Registrations are open until November 4th: https://www.lichess4545.com/lonewolf/register/ We hope you can make it for at least a few games!

Registrations for 960 Season 9 are open. It’s a 7-Round Chess 960 / Fischer Random Tournament with a time control of 15+15 with one game per week. You can register here: https://www.lichess4545.com/chess960/register/ Please make sure to join #chess960 after you register.

There is an upcoming blitz-battle today at 20:00 UTC. Registrations will open 1-2 hours before the starting time. To be notified when registrations open, make sure you join the #blitz-battle channel.

Series Season 8: In the Masters division @yago666 outplayed his opponents for first place @jfarah had a strong performance for second with @Jslavik coming in third. In the Candidate Masters division 1 @Brundd clenched first place @Tranzoo overcame his rival for second over @Numberman768 who got third. In the Candidate Masters division 2 @Isachess was clear first @pepepibote upset his way to second place and @Sesquipedalism became third.

Congratulations to Schaker ZHulu for winning Season 6 of the crazyhouse team league! @RapidVariants @AmelyPlaying @d4rkn3ss23 @MarlonC @Isachess @adande1 @Zher0 https://zh.lakin.ca/zhteam/season/6/summary/

Player of the season is Isachess with 9/10. This is also the second win in a row for RapidVariants and Isachess who were in the winning team in Season 5 also. Full recap and puzzles from the season will appear on the zhchess blogspot in coming weeks. Loch Zhess Monsters came second and TwelveTeam third. Well done to everyone for making an exciting and competitive season. 4th 5th and 6th all came essentially equal with only tiebreaks separating them.

Best results on each board:

#1 Jasugi99 (aka TwelveTeen) 6/6

#2 AmelyPlaying 8/10

#3 d4rkn3ss23 8/10

#4 marlonc 7/10

#5 Isachess 9/10

#6 colwem 8/10

#7 the_crocodile_hunter 8/10

The winner of the crazyhouse fantasy league was okei with 39 points but even this fell short of the 42 game points won by winners Schaker ZHulu which emphasises how impressive their performance was.

So you want to join the League? Who’s in and who’s out.

by @glbert and @somethingpretentious

Registration period is a boring time for most, but stressful for the mods. They sort through hundreds of registrations, checking for eligibility of potential players for next season (see rules I and VII for eligibility criteria) . The result of that busy time is publicly accessible to anyone checking the rosters, but let's take a quick look behind the curtain.

Figure 1. Registrations per season, stacked (A) and unstacked (B)

In this graph, I have grouped registrations as follows:

Repeaters are players who played the previous season
Returners are players who played in the league before, but not the previous season,
Newbies are players who played their first season,
Lurkers are players that were accepted as players for the season, but did not play a single game (data on these are only available starting with S6),
Rejects are players that were not admitted to play the season (data available only starting with S6).

If you are a veteran of the league, you are probably aware that as the season goes on, we too often run out of alternates. This is partially caused by the Lurkers, specifically the ones who do join slack but simply do not play any games – their numbers have been reliable over the seasons, but unfortunately they do fluctuate wildly from board to board. This makes it very difficult to judge how many alternates are going to be needed. If we misjudge these numbers, either alternates will wait for a ridiculous amount of time to get games, or players won’t have an opponent. Of course it is the individual alternate’s choice whether they want to accept games or not, so the mods will simply need to improve their predictions.

Secondly, we can see that the league has both a steady influx of new players, and regulars coming back; of course, many new players turn into regulars.

We can see exactly how much of a role the regulars play in the league, by looking at the Lorenz curve:

Figure 2. Lorenz curve of games per player

50% of the most active players account for about 85% of the games played in the league, and the top 10% of active players have played about 40% of all games. You may know this kind of curve mainly for checking out the equality of wealth distribution in countries, but it does work for our purpose as well. Our Gini coefficient is slightly below 0.55, so if games played resembled wealth and the league were a country, we would be similar to Japan.

Considering regulars are responsible for most of the games played – how is our player retention rate over the seasons?

Figure 3. Retention rates over seasons

As you can see, between about 60% and 65% of players return to play more seasons, with that rate being fairly stable. It does look like there is a slight downwards trend during the latest seasons, but in my opinion that is only because people who take breaks did not have the chance to return yet. If you compare this with the dashed line showing players playing the immediately following season, the current season to season retention rate is actually at a relatively high level, which makes me think that players from Season 15 and later will return – hopefully for Season 19, it is always nice to catch up with old friends and/or beat enemies you didn’t even remember you had.

Going back to Figure 1, there is one more thing that might be interesting: Rejected players.

Historically, players were mostly rejected for being being underage (Slack ToS), being previously banned for violating the respect rule (see VII.F), spamming, being abusive, or even trying to get additional accounts into the league. It may surprise you to hear that some are even in violation of rule I.A, or more specifically the “having a Lichess account” part. (As an aside, players that have no classical rating or only a provisional classical rating are currently not featured in Figure 1.)

Numbers for these rejections were always low, but when Season 7 was plagued by various cheaters (you can read about that in the ledger as well), the league mods decided to work with the Lichess mods to do various background checks on new players. These checks were implemented in the following seasons, leading to the uptick you can see in Figure 1. Later on, those checks were extended to regular players.

Most account refusals currently are for cheat detected games, artificial rating changes or using computer assistance. Note that these are checked not per account, but per player, i.e. if player A is marked as an engine user on his account A1, and registers for the league using his account A2, they are still rejected. The numbers in my figures are per account though, as these are the only numbers that are accessible to me. Players who get their account rejected will often try to get several different accounts into the league, so the number of rejections may be substantially inflated.

Still, for some of the later seasons, the number of rejections is close to the number of new players, so the bar for entering the league is much higher than many regulars assume. Additionally, the assumption that any marked engine user can simply create a new account and re-join the league is misguided.

It should be noted that there is a tradeoff here: As a player, I am not concerned about whether a person cheated on some other account, or years back. I am worried about people cheating right now in league games, and mostly only so far as it concerns my opponents in my own games. The Lichess mods make informed decisions however, that certain players have cheated before, and thus may be likely to cheat again. So we do restrict access to the league, and may even exclude people who may have no intention to cheat in league games. In fact, we do ban active players in the league all too often, who have bad scores and are thus unlikely to have cheated in league games, if we find they have cheated in less serious games or even on alternate accounts.

So how many cheaters do get accepted into the league, and how much of an issue is it?

Figure 4. Ineligible players per season.

The short answer is of course that we do not know. But we do know that people are caught cheating in league games, in non-league games, and on alternate accounts, and are thus ineligible for the league. I call them ineligibles. And for those we do have numbers. But I do need to emphasize again, that while many of them have cheated in league games, definitely not every single ineligible has cheated in all league games, and as mentioned some may even have cheated in none.

Looking at Figure 4, you can see that the highest number of participating ineligibles was during Season 8, but has steadily dropped since then. Of course, some participants of the latest seasons may turn out to be ineligibles later on, so I would expect those numbers to still go up significantly; having watched those numbers change over the seasons for a while now, I would expect the number for Season 18 for example to get a lot closer to 5% still, maybe even exceed it later on. I would expect that they do not go up enough for that visible decline to vanish.

The next question on your mind is likely: Who are those people? Are all of them on board 1?

Figure 5. Rating distribution of ineligibles compared to the rest of the league

To check out their rating distribution, I decided to create a split around Season 8, since that is when in-depth checks of registrations were first introduced. Before those the distribution of ineligibles was shifted noticeably to the right (Figure 5 left), with a bump around a rating of 2250, so many of these players would have been on board 1. In later seasons (Figure 5 right), the distribution of ineligibles is still slightly right-skewed, but with the exception of players below 1700, ratings of ineligibles are quite similar to the rest of the league.

So the number of ineligibles goes down, and they are almost as likely to be on a lower board compared to upper boards nowadays – but how do they actually score? How likely is it that you just lost to a cheater?

Figure 6. Scores by ineligibles.

Again, obviously we do not know, since even for earlier seasons it would be ridiculous to claim that everyone who cheated has certainly been caught. Figure 6 shows though (by superficial comparison to Figure 4, e.g. the 11% ineligibles in Season 6 scored 15% of the points in the league), that while ineligibles used to score above average, they do not do that anymore. The main reason for that is forfeits, obviously: If a cheater is caught during the season, this brings down their score heavily. But even excluding forfeits, ineligibles do not score much more than average. My interpretation of this is that heavy cheaters are likely to be caught more quickly, while the ones who manage to stay longer in the league have a more subtle approach and win fewer games; or maybe players who manage to stay in the league for longer simply start cheating later. And some of them, as mentioned, are unlikely to have cheated in the league at all.

In any case, if my interpretation is correct, this means that if you play two full seasons of 16 games total, you can expect to play about one game against an opponent who turns out to be ineligible, and who may or may not cheat against you. Most likely, they will not cheat drastically, so while they will play above their actual level, you will still have a chance to beat them. Or else you will have a forfeit win a couple of weeks later. Either way, it sucks, and we can all agree on that.

But we do try to work on this issue, and as you can see we have some success mainly by implementing preventive measures. And while more work is still needed, and discussion and input therefore is welcome, we do believe that most of the league is clean. So you do get to play relatively serious chess online from the comfort of your home.

Video

by @okei

Please feel free to join #lichessledger on slack if you would like to contribute towards the ledger in any way, or provide any feedback. Both are highly encouraged and appreciated. Thank you for reading.

Thanks to @glbert, @kostasvl, @kraaft, @okei and @Somethingpretentious.