Version 1.48f - please post feedback and suggestions here

[barmar]

Garbage In, Garbage Out

 

Nothing in, nothing out.

 

As I said, this doesn't have to be perfect. It only has to be better than throwing darts.

 

Do you think it's likely to make you select less compatible partners when you go to the Partnership Desk?

 

We don't currently take systems listed into account. But now, instead clicking on everyone in the PD, looking at their profiles to see if they look good, you can narrow it down to people with several stars. This will hopefully make your life a little easier.

[1eyedjack]

Since marking someone as a friend or enemy doesn't change your compatibility with them (see my previous post for how we use friend/enemy lists), the problem you describe doesn't occur.

 

atm, we're just experimenting, but we factor in ( unordered) profile-country, real location, language, masterpoints, avg adjusted points earned per hand, #people who mark u as a friend vs enemy , award symbol, starriness

 

I eventually plan to include ( for tourneys , anyway ) times played together , # friends in common

 

for the MBC ( main bridge club ) we're going to ultimately need a real/rich profile, i suspect.

 

Someone needs to make up their mind.

 

 

Friends and enemies: I mark JEC as a friend. I have never met him, I have no expectation whatsoever of playing with him or against him.

 

I suppose that if we have no intention of playing with someone, then it matters not what our compatibility rating with said individual states.

[barmar]

Someone needs to make up their mind.

 

Where is the contradiction?

 

If I mark you as a friend, it doesn't have any different effect on our compatibility than if I mark JEC as a friend. We look at the total number of friends and enemies that each person has, not whether those two specifically are friends/enemies.

[barmar]

The ratings are supposed to be symmetric, but I noticed a bug in this last night, regarding how we calculate masterpoint similarity. I'll be fixing that today.

 

This fix is in. If you still see cases where you and someone else see different compatibility with each other, let us know.

[scarletv]

Factors that could make sense for me:

similar completion rates for tourneys or main bridge club

similar speed in using the claim button

similar keywords in profile (better would be a catalogue of what is in scope and what you prefer)

similar avg points per hand

selfrating of skill

Possibility to reduce the compatibility rating (don't want to play with again)

 

Factors that do not make much sense for me:

number of friends I have or don't have

number of enemies I have or don't have

number of masterpoints (that mainly shows if you played the specific tournaments or not)

language not only for the bug being simply copied from the IP code

country not only for the bug being simply copied from the IP code

starriness

 

Instead of an increasing friend list I find assigning categories to players a very good way to see if I want to play with them again or not. I reduced the number of friends significant since that is available. Enemies I am using very rarely as it is not good when a TD is not able to talk to enemies. I can hardly believe that good partners should have a similar way to work with friends/enemies/categories.

[u4eni4ka]

I don't care about country, languages or masterpoints... seems the new feature is not useful for me...

 

If You find a way to estimate if my results with a particular player are higher then my average results and if this is true for the other player as well - I would find it useful. Also, if I'm compatible with a player, and He is compatible with a third player - it is very possible that I'm compatible with the third player too - so if I haven't ever played with a player this may give me some orientation.

 

For me my enemies are TOTALLY unacceptable as partners, so .... no need to be scored with more then 0 stars..

[WrecksVee]

So when can we sort by compatibility? This may be garbage in, garbage out BUT I would be curious to look at all the five stars to see if they seem to be a possible good partner.

 

My regular partner has not been on line and I eagerly await to see our rating. See the various Misadventures of Rex and Jay if you what to predict what might be expected IF that was the data source. We have been regular partners for nearly 10 years despite our differing views of many facets of the game. Given the factors listed for this I expect a high rating rather than the outliers shown in the Misadventures which are used by us to sort out partnership disagreements.

[FM75]

Speaking of ratings systems for compatibility...

I have noticed women's dance shoe advertisements in the left side of the screen lately. Surely you know that I am male. (I recently bought some birthday gifts online. Of course, I know why they show up there.) http://www.bridgebase.com/forums/public/style_emoticons/default/biggrin.gif

 

Sure, those advertising matches are not perfect for you, since I don't click on them. But they really don't bother me!

 

I am willing to bet that the day everybody thinks the compatibility rating system is good will be the day after everyone voluntarily stops using the windows client. It is probably also a safe bet that you will always have more complaints about it, than compliments.

 

I like the idea, and I have not seen any ratings that I found unusual. If I hit one, I won't lose sleep over it either. I will click on a woman's dance shoe advertisement. http://www.bridgebase.com/forums/public/style_emoticons/default/cool.gif

[jaywalk]

If I have no compatibility stars, what does that mean? No one will ever want to play with me! How do you get compatibility stars?

[Jakhammer]

This is great - another dimension to the bbo experience - and it seems amazingly perceptive!

[Mbodell]

I wonder if we could incorporate the trait of if you both understand what compatibility scores are and how they are not fixed for an individual but actually represent a rating of a pair into the compatibility formula.

[0 carbon]

Why stars? They already have too many meanings in BBO. Use smilies! Or teddy bears!

I think language compatibility should only be counted for countries where English is uncommon.

We can see MPs in the player bubble - don't use MP level as a criterion.

If either of you mark the other as an enemy, don't show ANY compatibility! It's likely that the enemier(!) won't invite and the enemieed will always be rejected. But enemy status just expresses incompatibility. It does not mean either is a jerk to all.

Categorized friends should be bumped up - you obviously have more interest in friends you have classified.

 

 

I'd rather see the rough average scores for the last 2 months or so as a partnership, eg "-3i 53%" You could even have it clickable to explore previous tourneys/hands.

 

For convention compatibility, check for any "words" in common in the profile. Ideally you would use a synonym list, eg, equating prec and precision. This also encourages people to put what they like in their profiles.

 

One extra thing you could also do is put a yellow (vs current black) 1-pixel box around players with whom you share a convention card. This would encourage more use of convention cards.

[0 carbon]

All the criteria above are possible indications of alikeness.

 

But isn't partnership success the main goal?

 

So why not look in your records to see what successful(good scores,repeated playing together) partnerships have in common Get a stats package like GNU PSPP and run a Logistic Regression to find out what successful partners have in common. Start with everything, and winnow out all the bad performers (correlation < .6) and weight the remainder according to the amount of covariance.

 

tOM

[1eyedjack]

If you have been flagged as an enemy by a large number of players, would that give you a high compatibility with someone else who had also been flagged as an enemy by a large number (all other factors being average/equal)?

[schi0]

I strongly support Carbon's opinion that PARTNERSHIP SUCCESS should be the main goal, or at least that is what one would expect of that ranking - a successful partnership with a casual partner.

 

while I am lost in the stats mumbo jumbo which seems pretty clever to me, I would suggest you take into consideration not only historical direct partnership success (A - B) but also an extrapolation like if (A-B) and (B-C) were successful, then (A-C) should be successful. Depending on computing power you can/would allocate you may go beyond A-B-C-D-....... and probably by 5th or 6th level any two players are connected.

 

I dont think that system preference should be a factor for casual partnerships since we tend to play a standard 2/1 with basic gadgets

 

obviously experts will have higher rankings overall given a higher success rate....

 

Compatibility is also about social skills so I suggest a mix of the two approaches

 

BTW, if you succeed to perform the regression and identify what really makes a partnership successful, please make it public

[john9999]

I agree with Jakhammer. I like it.

[barmar]

All the criteria above are possible indications of alikeness.

 

But isn't partnership success the main goal?

 

So why not look in your records to see what successful(good scores,repeated playing together) partnerships have in common Get a stats package like GNU PSPP and run a Logistic Regression to find out what successful partners have in common. Start with everything, and winnow out all the bad performers (correlation < .6) and weight the remainder according to the amount of covariance.

 

tOM

 

This is mainly oriented towards helping you select among people you don't already know. If you have an established partnership, you don't need us to tell you whether you're compatible.

[barmar]

So when can we sort by compatibility? This may be garbage in, garbage out BUT I would be curious to look at all the five stars to see if they seem to be a possible good partner.

In the Partnership Desk, click on the Compatibility heading and it will sort by them.

 

If you mean the MBC, we don't currently use the new compatibility rating to sort tables when you use List Interesting Tables. It does test for some similarities (that's how it decides which are "interesting"), but not all the factors that are included in Compatibility. Maybe this is something we'll add in the future, assuming it doesn't make the query too expensive.

[barmar]

If I have no compatibility stars, what does that mean? No one will ever want to play with me! How do you get compatibility stars?

 

You don't "get" them. As I've explained at least twice, compatibility is a comparison between you and the other player. The more things you have in common, the more stars you and they have when you look at each other.

[mycroft]

I would suggest that it be a bar graph instead of a star rating - that may make it clearer to people that it's "compatibility with you" not any sort of global rating. Sort of like the password complexity lines (although that is independent of user, I guess).

[hrothgar]

You don't "get" them. As I've explained at least twice, compatibility is a comparison between you and the other player. The more things you have in common, the more stars you and they have when you look at each other.

 

Arguably, you should list someone's compatibility with themselves as the highest possible rank rather than blank (which looks like the lowest possible rank)

 

Set this at five stars and you' be able to avoid a lot of bother.

[1eyedjack]

I think that you might want to choose something other than "stars" to indicate degree of compatibility.

Stars are already overloaded and folks associate this with skill levels and the like.

 

I think I can guess what symbol Lurpoa would suggest.

[c3cummins]

While I can't say that I completely understand the concept, I believe I have a better grasp now. Thank you for your input! Carol

[Oceanss]

This is mainly oriented towards helping you select among people you don't already know. If you have an established partnership, you don't need us to tell you whether you're compatible.

 

Yet, it's bit embarrassing having only half star with a good friend that one occasionally plays with, and does reasonable well.

[timouthy]

Two things: My own personal idea of a compatible partner is not just whether we can agree on a method of bidding (I already assume this to be true) but more importantly whether they will look at the results of a hand and decide whether they are satisfied with the result and adjust the bidding and/or play accordingly for next time. I also like people who are inclined to do this ASAP after the hand, but really anytime after the hand is also fine. If you could assemble this info. for me in the next version of your compatibility algorithm, say in a week from now, you can assign 5 stars to anyone who says they are like this, no matter what else they say about themselves. We all make "mistakes" playing this game and not all are entirely foreseeable. The goal is to always be looking to improve and the faster you improve, the better you perform.

 

Next: I hope you never get rid of the friends/enemies list. I don't care one whit that the term enemies has a negative connotation to it. There are all sorts of reasons to put someone on the enemies list. I could list them here but don't get me started. On my list I always include the reason they were put there, so I know why. I could be wrong but I don't think most people's careless/thoughtless/rude behavior changes much over time. If I could copy my list and disseminate it, with the reasons to avoid someone on BBO, I would think it would save others lots of time and avoided insults.

 

One more next: You need more skill levels. I am between advanced and expert and there should be a skill level for this and probably one between intermediate and advanced as well. Since you didn't have one, I had to get creative and name myself this way. I now can get onto most expert tables and compete there where before I wouldn't have been allowed.

 

sincerely,

Adv exp