I can't play 5 pointers

[boop]

The other day I mentioned on Fibs that I'd had a terrible day playing 5 pointers against gammon bots and could anyone give me any ideas on the difference between 3 and 5 pointers as I generally only play 3 pointers.
I said that I'd probably lost about 20 matches in a row and got caught up in a discussion about how accurate my memory was. (I'm pretty sure it was at least 12 - A LOT - and probably more)

The next day I went back to playing 3 pointers and got my rating back up to 1870 ( i've been hovering around 1900 for the last 6 months)

well today, 2 days later it's happened again and i've been counting. I lost 6 gammonbot 5 point matches, won 1, and then lost a further 13 in a row!!! I want to smash, kill and burn things ...

so ... what's occurring? ..... is the cube really different? ..... anything else?

[KissMyAss]

I find I hit patches like that with three pointers.  Look at your stats.... see where your highest percentage of wins are, and play to that for a while. 

I have a higher chance of winning one pointers than I do three or five or beyond.  So I play one pointers with bots, and the larger matches with people (usually during tourneys). 

You already know your rating climbs at three pointers, so play those for a while, build yourself a cushion of rating points you can afford to lose, then try to rehammer the five pointers.   

I don't know if it makes any difference or not, but to me it does.  My other technique is... if I win against one bot, I will play that one again.   If it beats me, I move on to a different bot.  Once i've lost to it, I don't re-play it again for a while.  Just to break its run of beating me.  Like I said, no idea if that actually makes a difference, but that's how I roll. 

Hugs

[dorbel]

Imagine that you have reached the final of a major tournament. Your opponent is one of the ten best players in the world, Falafel let's say, or Mochy, or Sander Lyloff, it doesn't much matter which. He is highly motivated, partly by the prize money, partly by pride. Even more importantly, he is in the habit of playing well, always considering every position on its merits, thinking ahead, planning. He'll often take half a minute to think what to play, more for an important cube. How will you approach this match? Ideally you will be rested, fed, sober and ready to give every roll your undivided attention.

How is this relevant? Because a Gbot is about as good as any of the above, plays considerably faster and can do so for a week without sleep, food or drink. How will you approach this match? Probably with the radio on and simultaneously making coffee and chatting to somebody on the phone! You'll be monitoring shouts and wondering if you can fit in another match before the tourney starts. You may be tired, you may have consumed an adult beverage or two. Worst of all, you will play far, far too quickly and even if you start playing slowly and carefully, you won't be able to keep it up, because you are just not in the habit of playing like this. What is your win ratio in five point matches likely to be in these circumstances? That's for you to say, not me!

What to do? I suggest playing better. How to do that? Play fewer matches, play more slowly and above all, train yourself to give your undivided attention to every play that you make in every match that you play. This requires brain stamina and that is only acquired with practice. A carelessly played match "while I drink this coffee", "before I take the dog for a walk", "while I'm chatting to Auntie Dolly on the phone" does untold damage. When each match is over, analyse it and think carefully about your biggest errors and how you came to make them.

Lastly, yes there are considerable differences in cube play between three and five point matches and as you may suspect, three point matches are much easier to play. The cube decisions are easier of course and the match is shorter, so less of a stretch for your attention span.

Playing good bg means making the maximum use of what you already know. Studying more won't lead to more wins unless you can bring what you have learned to the table and that is done by acquiring good habits. Everybody below world class level, 99% of us, can improve their play and their results just by better preparation, concentration and above all the will to do it. Without these, even the most talented will never achieve their potential. Even with all that, you will still sometimes lose ten matches in a row, just less often! That's BG.

[boop]

thank you both for your answers. There are some very good points made and I shall take these onboard. especially taking my time each move.

kma - as far as i'm aware the gbots are all the same program so I just pick the highest rated available and play. That way I lose the least points or win the most
... hmm ... that's a point - are they all accessing the same neural net or do they have their own and continually improve?

[ah_clem]

I'd recommend analyzing and saving all your matches.  Both XG and gnu allow you to keep a database of saved matches so you can query them to see how you are progressing.  XG's db is much simpler to use, so I'd recommend it over gnu.

With your matches in the db, you can easily see whether your losing streaks are due to poor luck or poor play (or a combo of the two).   XG will give you summaries of your performance rating (PR) and luck grouped by match length, so if you really play worse in 5-pointers it will be obvious. The first thing to do is to determine whether you really do play 5 pointers worse than 3 pointers.

If so, then it's time to investigate further.  Are you losing equity because you don't understand certain match scores?  5-pointers are long enough that there are several tricky scores; 3 pointers don't have any tricky match scores (or should I say that there are few and well understood)

Perhaps you are making checker play errors that would not be errors at a different score.  If so try to identify the trouble match score and learn a bit about it.


////////////////////////////////////////////////////

are they all accessing the same neural net or do they have their own and continually improve...

All the gammonbots are gnubg with some pearl wrapped around it to interface with FIBS. My understanding is that the neural net is "static" within a version of gnu and doesn't change (learn) as it plays. Newer versions of gnubg may undergo additional neural net training and be stronger.  AFAIK, all gammonbots use the same version of gnu and the same settings and therefore can be assumed to be identical.  Playing the highest rated one is a good strategy to maximize your rating, but be aware that this may mean that you are over-rated.

[Zorba]

Good points made by all. If it's really the fivepoint match that bothers you, it could well be because that one probably has the most games with tricky matchscores (on average). Kit Woolsey's article explains it really well, it's a long read but well worth it: http://www.bkgm.com/articles/GOL/Aug99/fivept.htm

[garp_02]

How did you get to 1900 would be a more pertinent question 

[dorbel]

Probably by avoiding five point matches! If you stick to one or three point matches, your error rate and thus your rating is better than it would be if you played longer matches.

[boop]

my guesstimate of previous playing habits

85% 3 point matches against highest rated available gammonbot
5% 1 point match against highest rated available gammonbot
10% 5 or 7 point against humans - various ratings although if I invite they would be 1850 or above

I am now for the forseeable future playing 5 pointers against gammonbots chosen from the middle of the list of available bots ... while keeping the above 10% of humans.

I think that article will be very useful - thanks ... I know match equities have not been part of my game and should definitely bump my rating up a bit.

I do try to learn from my match analysis on GNU after every single match and keep them in a folder so if GNU can tell me more about my play over a number of matches that would be fantastic.

I've found this quote

http://www.gnubg.org/documentation/doku.php?id=introduction#3

"Record keeping of statistics of players in games and matches using relational databases (SQLite, MySQL, or PostgreSQL) and Python."

... but can't find a page that describes how to do it - can anyone provide a link?

boop

[dorbel]

Or splash out $50 on ExtremeGammon and have it all done for you. You won't ever regret it. (An unsolicited testimonial)

[garp_02]

I have to agree with Dorbel. XG is a brilliant aid to improvement. I recently purchased it and stopped playing on Fibs for a month or so. Since returning to Fibs, only playing Gammonbots and a few FLG matches for 7 points, my Fibs rating has improved by over 100.

Well worth the dosh

[boop]

is there a mac version yet?

[dorbel]

No Mac version yet, although as you can have a version for your i-phone and i-pad, I dont know why.
The thing with XG is that you can track your progress in improving your error rate. Judging how well you are playing by your fibs rating is not very helpful because although better play will mean that your average fibs rating will be higher, fibs ratings vary through a range of at least 200 even with consistent play!
Looking at your error rate is much more helpful. I have improved mine by a full point in the last few months. XG's compilation of data is comprehensive and very user friendly.

[stiefnu]

No Mac version yet, although as you can have a version for your i-phone and i-pad, I don't know why

Apparently a Mac version is due to be released next year.  Meanwhile, the 'computer' version (PC only) costs $59.95 while the app versions (Mac & PC) are a mere $10 or so.  Anyone know what the expensive version does that the app does not - extended rollouts, maybe?

[ah_clem]

"Record keeping of statistics of players in games and matches using relational databases (SQLite, MySQL, or PostgreSQL) and Python."

... but can't find a page that describes how to do it - can anyone provide a link?

Here's a good intro to the SQL language which you would use to interact with the relational database

http://www.w3schools.com/sql/sql_intro.asp

I've never been successful at finding documentation of the data model used by gnu, so even though I'm very proficient with SQL I could never figure out how to use the gnu built in database.

XG makes it easy - no need to know anything about SQL. 

[boop]

right - i'm not that interested in backgammon to get into sql!!

so I downloaded the demo of xg on my iPhone and it gives me a rating after each match and the number of blunders.
if I pay the £6.99 to go pro, will it then analyse batches of my past games and give my insight into my kind of errors as it seems the PC version does?

[boop]

i bought it anyway as i like it ... will report back 

[garp_02]

Well done Boop.

Not sure what the app version is like but the PC version is very good.

As Dorbel rightly says, the ability to check, and analyse, your errors is a much better means of checking your progress than simply how your rating improves. I have found out (if not already known) that my checker play is consistently pretty good but my cube play can vary quite dramatically. So I am trying to use XG, at the moment, to improve my cubing (giving and taking).

The other thing which I notice with my 500+ games with XG (Champion Level) is that the vast majority of matches are won, by either side, with the help of luck. More of my games are won by luck than XG's, but I guess that it shows that when two people of similar ability play, the result will, more often than not, be settled by the luck of the dice.

There are many people on here who have much more knowledge on this subject than me, but just thought I'd share my thoughts.

[boop]

well extreme gammon on the iphone give you these stats

Elo rating, Experience - number of games

Level of play:
overall : PR number
checker play : PR
Cube : PR

Plus the percentage of matches won overall which is also broken down into matches of different lengths.

Matches can be exported via email so I can load them into GNU for analysis.

overall a good £7 spent.

I've boiled Kit's 5 point matches page to bullet points and it's being very useful. lots of things in it that have cleared my thinking including having an initial strategy which now makes clear Dorbels Gammon Go , Gammon Save columns in his beginners series...

I'm still slipping down the ratings ladder though now that i only play 5 pointers- lots of really bad decisions being made - many xg would describe as "distracted" games so lots things to work on ....

[dorbel]

Don't be discouraged, to mix study and play is extraordinarily difficult, which is why most tourney players do nothing but play for at least two weeks before a big tourney. Since I stopped writing dorbeldaily, which of course involved a lot of analysis, my PR has improved a lot. Keep going, in the end it will be worth it, not for a better rating, though you should get that, but because the game gets more interesting when thought is applied to it.