Chess Playing Style - New Indices

There have been several discussions about member feedback e.g.

sportsmanship-meter

sportsmanship-rating

i.e. after you’ve played against someone, you rate them on the playing experience. The thinking behind it is that members who misbehave (e.g. by chatting abusively) will get ‘gunned and shunned’. Several well-known sites operate such schemes, most notably eBay and Amazon.

Many people find the concept of peer review deeply flawed.

• They assert that is untrustworthy - since people are likely to subjectively rate their friends positively and their ‘enemies’ negatively;
• that it is intrusive and divisive – and hence out-of-place in a friendly, game-playing, social-networking community;
• that it is a disproportional response to the relatively limited problem of abusive, rude or childish behaviour (and for which blocking and/or reporting) works satisfactorally);
• and that there can be no agreed standard against which to measure unacceptable behaviour.

It is true that, members’ profiles already contain a wealth of indicators as to likely behaviour.

• Rating – a high rating usually means that someone is deadly serious about chess!
• Member points are a measure of activity in the forums and may correlate with likelihood of chatting during a game.
• Time/Move – shows whether someone plays quickly or slowly
• Current Games – if high, can mean that someone won’t spend long thinking about each move
• % Timeout – if non-zero, indicates that the person either has more games than they can manage or that their connection is unreliable
• Last Online – sheds light on whether someone is online daily or weekly

Some of these statistics can be used as filters on Game Seeks and Tournament creation.

The set of Groups to which a member belongs is publicly accessible – as is the set of trophies they have received (unless deleted). Both of these lists provide an insight into a member’s views and interests.

In addition, the profile itself contains whatever information the member chooses to reveal. The voluntary Extended Personal Profile provides a structured approach to organising this description. Essential Chess Type is another voluntary scheme for providing a quick overview of chess personality. Most members will not see the point of them.

Forum topics initiated or contributed to give a strong pointer to a member’s attitudes.

Judging from the volume of discussion in the forums, there are three aspects of chess playing where opinions as to what constitutes good or bad form - are particularly polarised:

1. Acknowledgement – whether your opponent says ‘Good Luck’ (or equivalent) at the start of a game, and ‘Good game’ (or ‘Thankyou’) at the end of the game.
2. Assistance – whether your opponent is comfortable with using Correspondence Chess facilities, in particular, Game Explorer and the Analysis Board
3. Resignation – whether your opponent plays until checkmate, or resigns

It is clear that a subjective feedback scheme for rating members on the basis of how much you enjoyed playing them - would not be universally welcomed. Nor would it express sufficient useful data in a single index. Many members would opt out, and others would be reluctant to express an honest opinion in case of reprisals.

However, these three aspects could (to varying degrees) be measured objectively and automatically. The statistics would be of interest both to members themselves, and to the wider community.

A measurement related to Acknowledgement would be: Average Length of Chat / Game (in characters).

For Assistance: percentage of games in which either the Analysis Board (easy to do) or Game Explorer (difficult to do) was used. {The latter would be much easier if Game Explorer were launchable from the game.}

For Resignation: percentage of lost games that were resigned.

In summary, these modest additional statistics would make members’ playing styles more transparent, and would provide a more informed basis for people to choose opponents in a non-judgemental way. Ideally, the statistics would be made available as filters on Open Seeks and Tournament creation – alongside the existing metrics.

I just want to play, I dont care if they talk or not, resign or not, use assistance or not.

As long as they are not cheating, and are not extremly abusive, I'll play with anybody if I feel I can handle more games.

The point about resigning is that:

A) some people DON'T CARE whether their opponent resigns or plays until checkmate

B) some people get annoyed if their opponent DOESN'T resign

C) some people get annoyed if their opponent DOES resign

The proposed metric will enable the people in group B to select opponents who might resign, and the people in group C to select opponents who are unlikely to resign.

I use eBay extensively as a buyer and I find their feedback system works extremely well. The first step is to rate the transaction as either positive, negative or neutral. You can then add up to 80 characters of explanation.

The second stage is to add detail for specific aspects:

• How accurate was the item description?
• How satisfied were you with the seller's communication?
• How quickly did the seller dispatch the item?
• How reasonable were the P&P charges?

   in terms of 1 to 5 stars.

As eBay points out: "These detailed seller ratings are anonymous, so please feel free to leave honest ratings about your buying experience."

If you give a low rating for the item description, eBay requests further information ...

Why weren't you satisfied with the item description?

• Item was not received
• Quality did not match item description
• Item was damaged
• Item was a counterfeit, replica, or an unauthorised copy
• Wrong item
• Other

As a buyer, if I see that a seller has less than 100% positive feedback, I read the comments left by other buyers and examine the detailed feedback to make a judgement whether to trust the seller or not.

eBay has a further refinement. It shows you the feedback profile for any member over the short, medium and long term.

This enables you to determine whether a seller's satisfaction rating is improving or getting worse.

What do you consider to be the problem with eBay's feedback system?

I don't see it that way.

The people who want to use Correspondence Chess facilities on this site are presumably well aware that they are perfectly entitled - even encouraged - to do so.

Some members, though, are unaware that we are all playing Correspondence Chess, or that the conventions of Correspondence Chess are not the same as those of Over-The-Board chess.

Other 'chess-purist' members wish we were all playing OTB chess, and some disapprove of CC conventions.

Currently, the best route for finding opponents who share a chess-purist outlook is to join a group such as Circle of Trust - which puts its members ON TRUST to avoid using CC facilities.

If this ASSISTANCE metric were available, you could just look at someone's profile to see whether they had used CC facilities. There would be no need to rely on trust or honour.

If someone uses an external opening database (which they are perfectly entitled to do), it would be a simple matter to provide a tickbox for them to manually record the game was assisted. No member here that I know is ashamed of using Opening Databases or other permitted facilities, or wishes to conceal the fact.

The point about chatting is that:

A) some people DON'T CARE whether their opponent chats or not

B) some people get annoyed if their opponent DOESN'T chat

C) some people get annoyed if their opponent DOES chat

The proposed metric will enable the people in group B to select opponents who are more likely to chat, and the people in group C to select opponents who are unlikely to chat.

These proposed metrics are not going to be of any use to you. Sorry.

Art I think you've made some good points here regarding assistance and resignation, however as for acknowledgement, I fail to see how a raw character count would be an effective measure of any ones chattiness. Some are quite adept IMHO to LOL W/O using  a big(see large would've been more letters) # of characters.

Don't forget this is referring to in-game chat - not to forum posts. You're right: a simple character count is far from being an accurate measure of Acknowledgement - it's more a measure of Chattiness. Nevertheless, it serves as a unobtrusive indicator. It distinguishes between the players who say something and the players who say nothing.

Paul, your posting seems to occur twice. You could edit the repeated part out.

Paul wrote: the first question, before the rating could be: do you want to chat, acknowledge your opponent or do you prefer to just play the game.

I wasn't proposing to ask any questions - merely that the system would automatically analyse the situation. Posing questions before you can start a game would be intrusive and unwelcome.

 Resignation – whether your opponent plays until checkmate, or resigns 

Paul wrote: Not a clear cut situation here as I try to help new comers and some never see the ending until checkmate happens or they want to see how I will do it, just to learn end game. Possibly the rating here should be, arbitrarily set at say 1700 rating as under this number many opponents may not see the checkmate.

Again I was proposing that the system simply count resignations just as it currently counts wins, losses, draws and timeouts. A high resignation count is neither good nor bad per se. Those players who prefer resignation-prone opponents can select members with a high resignation count. Those players who prefer play-until-checkmate opponents can select members with a low resignation count.

  Assistance – whether your opponent is comfortable with using Correspondence Chess facilities, in particular, Game Explorer and the Analysis Board 

Paul wrote: A good question to ask and perhaps if the opponent is reluctant to use such tools one should comply or resign the game.

Once again, I wasn't proposing to pose any explicit questions, but to leave it to the system to determine automatically whether tools were used. As usual, use of a permitted tool is neither intrinsically good nor bad.

Not a lot of algorithmic coding - just need to note the three bits of information: (1) whether game was resigned, (2) length of chat, and (3) whether Analysis Board or Game Explorer was launched (assuming GE was launchable from within the game) or Opening-Database was explicitly ticked.

Most of the work would be in displaying the results. The statistics page would have to be reformatted, for a start.

Clearly a waste of effort in the light of more pressing features and fixes.

Some message boards have a feature where you can thank a person for their post, and then see the number of thanks that that person has received just below their avatar.  In general, the people who have been thanked quite a few times are more likely to be helpful than those who have never been thanked.  Perhaps chess.com could implement a system where the number of times you have been thanked for a game appears on your profile page.

Would the average number of trophies you receive per game be a reasonable estimator of this?

Yeah, that would be useful.  Sure.

And the average number of trophies you give per game would be revealing too.

The number of times thanked or the average number of trophies received/given would be additional metrics alongside the proposed three i.e. chattiness ('Acknowledgement'),  purism ('Assistance' i.e. happy to use GE/AB) and 'Resignation tendency'. 

(Interestingly, an opponent resigned one move from mate in a recent game. However, I had only seen the checkmate myself one move earlier, so arguably I should have been penalised under the windfall-tax rule.)

In the following game, that maestro of attack Alexei Shirov was playing Canadian GM Kevin Spraggett.  In the final position, Shirov fell into a deep think, trying to decide how to continue the attack.  Suddenly, Spraggett extended his hand in resignation.

"What?  Why?" Shirov asked, puzzled.

"It's mate in two," Spraggett replied.

I see mate in three:

1. Ng6+ hxg6 2. Qh3+ Qh5 3. Qxh5#

Nice finish, even if it wasn't planned.

2. Qh6#