What are the Daily Quiz Recaps?
Every week on my forums I post a set of 5 problems semi daily or 4-5 times a week. These will always appear with a subject line of "Daily Quiz" followed by that day's date. These positions are taken mostly from my own play against the computer though when I do attend live tournaments I end up amassing a good bit of positions from there and share them through the daily quizzes. Ideally if you sign up for the recaps I would prefer if you also visited the forums and posted your own thoughts on these positions when you have the time. The more people that reply to the recaps the more information I have at the end of the week on which positions I should cover. If everyone who replied on the forums failed to grasp a major concept that would be a problem I would try to include whereas maybe I thought another problem was worthy of inclusion but all who replied got it right so I would omit that one in favor of another more difficult problem.
At the end of every week I write a recap which is more or less an article where I select 5 positions from the previous week and explore them in depth. I've included a few examples from former recaps at the end of this page. I provide rollout results for all problems from that week using eXtreme Gammon and very high level settings. Then I tackle the 5 selected problems from that week providing my own insight, the computer results, variants on the position to help pick up on the key factors and anything else that I think is pertinent. You will receive all the XG files of the rollouts from that week, the 'article' on the 5 recapped problems in HTML form, and also the 5 positions I selected in another document without analysis so if you didn't have a chance to look at the problems during the week you can view them and make your own decisions before reading over the recap. These will be sent to me via the email address you supply me with. They will normally be sent out over the weekend, any time between Friday and Sunday.
The cost of this is $10 a month or $100 a year. If you are a student of anyone at The Backgammon Learning Center I offer the discounted price of $50 a year. State when making payment that you're a student of Phil, David, Perry, or John and I will verify with them. I have also extended that price for students. There are various ways you can make payments, the easiest being via Paypal. The email address I have attached to Paypal and where you'd send payment is the same I use for all my email correspondences: checkmugged@yahoo.com
If Paypal doesn't work for you sending a check, money order, or cash (at your own risk) is fine by me. Email me and I'll be happy to get back to you with my address and real name if needed. You may also retroactively start your subscription from any date you want. For example, quite a few people have joined along the way requested to be included from the beginning and this is no problem. I can email you the files from any given week or every week if you like. If you ever have any comments, questions, etc... about the daily quiz recaps or the problems within' you can feel free to email me.
There has only been one long review (thanks Timothy) and a few short reviews of my recaps but I offer them up below.
Timothy Chow
mtuhtan
KenB
Bill Calton
Now let me offer you up a few problems from past daily quiz recaps.
Number 5 from 5/9
|
| | Analyzed in Rollout | | No Double | | Player Winning Chances: | 68.31% (G: 27.12% B: 0.16%) | | Opponent Winning Chances: | 31.69% (G: 6.40% B: 0.34%) | | Double/Take | | Player Winning Chances: | 69.83% (G: 28.17% B: 0.23%) | | Opponent Winning Chances: | 30.17% (G: 5.97% B: 0.31%) | | | | Cubeless Equities | | No Double: | +0.366 | | Double: | +1.236 | | | | Cubeful Equities | | No Double: | +0.731 (-0.242) | | Double/Take: | +0.973 | | Double/Drop: | +1.000 (+0.027) | | | | Best Cube action: Double / Take | | | | Rollout details | 5184 Games rolled with Variance Reduction.
Dice Seed: 21140344
Moves and cube decisions: 3 ply | | | | Confidence No Double: | ± 0.005 (+0.726...+0.736) | | Confidence Double: | ± 0.011 (+0.962...+0.985) | | | | Double Decision confidence: | 100.0% | | Take Decision confidence: | 100.0% | | | | Duration: 39 minutes 14 seconds |
|
This type of problem has given me trouble in the past. Your opponent has you on the bar, five point board, and it's the rest of the elements that decide whether the position might be a take or you'll do what I used to, take massive drops. The point that is open is obviously a factor but in general I had problems with a problem like this, the two point specifically being open. In this problem our opponent has no builders in direct range to hit loose or point on us if we enter on the deuce point immedaitely but even this is somewhat of an illusion. Our opponent is on roll and of course he is likely to unstack his 10pt. Given the results we can imagine that even if the builder was back a pip but diversified by being on the 11pt this would be pushed into a pass. Let me show you a few variants and see how you do.
Well, did I con you in to taking any of these? The race is relatively close in all of them, our opponent has no men even close to the zone to continue the attack should we enter quickly, our offensive structure is decent and our opponent has yet to free his back checkers. Massive passes, all of them. Short rollouts are provided in the files I attached if you want to look at them. The first two are passes by ~.400+, a quadruple whopper with all the fixins, and the last one which I'm sure the vast majority was lulled into taking as it's the best of the set rings in just under a .200 double blunder to take. Most people overestimate their winning chances and underestimate the gammons they lose resulting in these take blunders. You have to ask yourself how likely it is you enter that checker in 1 roll? 2 rolls? 3 rolls? And when you do what is going to likely happen? You'll probably be attacked again and forced to again enter quickly or suffer the consequences. For reference you're a slight favorite (~52%) to enter 1 checker against a five point board in two turns. That means in two turns you will not have entered the checker ~48% of the time. Brutal. For reference:
Probability of Entering One Checker in:
Points Open 1 Roll 2 Rolls 3 Rolls 4 Rolls 5 Rolls
1 30.55% 51.77% 66.51% 76.74% 83.49%
2 55.55% 80.24% 91.22% 96.10% 98.26%
3 75.00% 93.75% 98.43% 99.61% 99.90%
4 88.88% 98.76% 99.86% 99.98%
5 97.22% 99.92% 99.99%
6 100.00%
Our original position is better than all the examples in that our forward position is more advanced thus allowing us to not lose so many gammons and win slightly more games than in the variants. I still dropped this OtB. I've become gun shy about these positions because as you can see it takes so much for it to be a take and you could easily slip up and super blunder.
Number 1 from 6/9
| | Analyzed in Rollout | No redouble | Redouble/Take | | Player Winning Chances: | 75.78% (G:0.00% B:0.00%) | 75.75% (G:0.00% B:0.00%) | | Opponent Winning Chances: | 24.22% (G:0.00% B:0.00%) | 24.25% (G:0.00% B:0.00%) | | Cubeless Equities | +0.516 | +1.053 | | Cubeful Equities | | No redouble: | +0.900 | ±0.003 (+0.897..+0.903) | | Redouble/Take: | +0.900 (-0.001) | ±0.004 (+0.896..+0.903) | | Redouble/Pass: | +1.000 (+0.100) | | | | Best Cube action: No redouble / Take | | Percentage of wrong pass needed to make the double decision right: 0.7% | | Rollout details | 2592 Games rolled with Variance Reduction.
Dice Seed: 50438717
Moves and cube decisions: XG Roller+
Search interval: Huge
| | | | Double Decision confidence: | 61.4% | | Take Decision confidence: | 100.0% | | Duration: 13 minutes 51 seconds |
|
We end the week on yet another 'easy' position. I chose these positions because the people on this list should know these cold. Small tweaks in these positions make a big change and if you have the reference position in your memory bank and have messed around with it before then when positions like the one above come up, and it actually did come up OtB, you will have no worries. So first to the reference position most people were referring to:
| | Analyzed in Rollout | No redouble | Redouble/Take | | Player Winning Chances: | 78.10% (G:0.00% B:0.00%) | 78.07% (G:0.00% B:0.00%) | | Opponent Winning Chances: | 21.90% (G:0.00% B:0.00%) | 21.93% (G:0.00% B:0.00%) | | Cubeless Equities | +0.562 | +1.123 | | Cubeful Equities | | No redouble: | +0.965 (-0.035) | ±0.002 (+0.963..+0.967) | | Redouble/Take: | +1.002 (+0.002) | ±0.002 (+1.000..+1.005) | | Redouble/Pass: | +1.000 | | | | Best Cube action: Redouble / Pass | | Rollout details | 10368 Games rolled with Variance Reduction.
Dice Seed: 8398984
Moves and cube decisions: XG Roller+
Search interval: Huge
| | | | Double Decision confidence: | 100.0% | | Take Decision confidence: | 95.5% | | Duration: 16 minutes 37 seconds |
|
The latest versions of XG allow us to have reference positions for these end game situations that are more precise than ever in a reasonable amount of time. This reference position has been known aas borderline but a take for as long as I can remember now. The fact that it may be borderline but a drop won't make much difference out in the real world but I like things as precise as I can get them. OtB you should of course take against any human if the exact position were to come up because they likely don't know how to play it out as well/aggressively as the bot. Remember this is a borderline take/drop because we have recube vig. Some people thought this position was borderline and the winning chances split was 75%/25%. Not true and may one day be of importance to you in match play. It is only a take for money because of our recube vigorish.
So let's say we did know the reference position cold, what now if we are considering the cube? We know it's a take since the score doesn't affect the take point, should we still double? This is a position I believe it is of utmost importance to double against a human. Your distribution is worse than your reference position and you have 4 immediate blotting numbers but you should be able to figure out that these details aren't likely to make it a no double by much if at all so by cubing you give your opponent a chance to make a sizeable error. I think it's much more likely a human makes the error by passing now than by taking after the next sequence when it goes well for you and he has a clear pass.
Why might your average player get this wrong if it's such a common position? The main reason is they just don't know the position regardless of how common it may be to you. Seeing people misplay the second roll of the game after knowing they've seen it for the last 30 years and the best play has never changed has convinced me never to underestimate a human by overestimating him. Even if they have a vague idea that this position might be a take they may talk themselves into passing due to the score. You and I both know the score has basically no effect on the take point here but Joe Donkey doesn't, put him to the test. The cube, especially at an elevated value, is often a weapon that people don't like to see used. They see this position and start mumbling to themselves about losing 4 points and there's no play left in the position except hoping to get lucky. They came to a backgammon tournament to play, they don't want to shorten the match that way, they'll pass this and even if it's a 'small take' they won't mind. They came to play.
Now we've covered the theory of why you should double this against your average every day player, what if you were playing Neil Kazaross, Falafel, or even the mythical Stick? The double is of even greater importance! This may seem counterintuitive at first since all of us will know what the position is worth and the theoretically correct cube action. However, that cube action assumes equal opponents and I'm assuming that this match up is not equal. In fact if you want proof and you have GNU you can load the jac100 match equity table (settings >> options >> match >> load) and see for yourself what the proper cube action is of our reference position with a meager 100 elo point difference. 100 elo is worth about 3 on the PR/ER scale so if you take the 3 players above and say they average about a 3 PR do you average a 6 PR or worse? If so, that MET will give you an idea of what a huge pass this position now is for a stronger player against you. The reason being that a stronger player doesn't want to shorten the match with big cubes. The stronger player certainly doesn't want to play out positions where there's not a lot of play left like this one. The stronger player's recube vig should he get back in it is less than normal since he's not playing an equal opponent. By cubing these stronger players you will also often get a mega blunder take in return because they 'know' the position is a take. A lot of people talk about adjusting according to the opponent, few people do it in such positions.
I'm getting well off topic now but you can stop reading any time you like =) In a lesson I was giving just this week the position below came up. I'm the top player, my student is on roll and doubles me in this position, what to do?
I have to say I was ecstatic about my estimates and OtB actions once I confirmed them. I took a minute, ran through the numbers in my head, and passed. I had estimated a 65/35 winning percentage split and a 20% difference in gammons. We had already figured out the raw and adjusted take point as part of the lesson so it was good we got to such a decision as I was able to toss in the gammon value to boot and show how that affected my take point. So I hit the 'pass' button and XG pops up my PR that game as a whopping 23! That's right, Michael Jordan. I quickly told my student that the bot is a donkey and doesn't understand certain things and that afterwards I would show him how I played a 1 PR instead of a 23 PR. We finished up the match, I opened up GNU, swapped the MET again and boom, even with a 100 elo difference this is a blunder to take. XGR+ has this as a double blunder to pass or a .164 mega whopper for anyone interested. GNU with the jac100 loaded has it as a 1.1 pass. I'm not sure how strong this student plays on average but I figure the average difference is at least 3 PR so you can imagine what a sick blunder it would be in practice to take. Without getting into it I will note there is more to it than that as this position may not be the easiest to play for my opponent if I manage to get the backgame but I still don't think that's enough to overcome the monster pass I assumed it was AtO.
Number 5 from 6/16
| | Analyzed in Rollout | No redouble | Redouble/Take | | Player Winning Chances: | 79.32% (G:0.00% B:0.00%) | 79.32% (G:0.00% B:0.00%) | | Opponent Winning Chances: | 20.68% (G:0.00% B:0.00%) | 20.68% (G:0.00% B:0.00%) | | Cubeless Equities | +0.586 | +1.173 | | Cubeful Equities | | No redouble: | +0.864 (-0.136) | ±0.001 (+0.863..+0.865) | | Redouble/Take: | +1.135 (+0.135) | ±0.001 (+1.134..+1.136) | | Redouble/Pass: | +1.000 | | | | Best Cube action: Redouble / Pass | | Rollout details | 5184 Games rolled with Variance Reduction.
Dice Seed: 61444079
Moves: 3-ply, cube decisions: XG Roller
| | | | Double Decision confidence: | 100.0% | | Take Decision confidence: | 100.0% | | Duration: 1.7 seconds |
|
These problems can be messy if you don't know what you're doing the best you can do is eyeball it and says "looks like a 4 roll v. 4 roll so I guess I take?" That's fine if you have nothing better at your disposal but the point of you subscribing to these recaps is hopefully to have something better at your disposal. I give to you again the effective pip count. In the following link you will find the absolute basics you need to get started with the effective pip count. You can't progress without knowing some things cold and then you can start to work with the easy cube formula and improving on your estimation of epcs. Effective Pip Count Basics
Here the bottom player's epc is easy. We start with it being a 4 roll position (refer to link if needed) so 7(4)+1=29 is the start. Now we aren't yet finished. We need to adjust a little bit because this isn't a pure N roll position. How to make these adjustments is at the bottom of the link I provided. In this example we have an even number of checkers with each checker being worth on average ~2 pips (17/8). If we consult our chart we see that we therefore need to add 1 to our previous epc count bringing our final epc count to 30. If we check with a bot the actual epc count is 30.4 but that's pretty good for a human.
What about the top player, can we epc for him!? Yes! First let's start with his checkers on the ace point and get them out of the way. They alone are a pure two roll position so 7(2)+1=15. Now to the outfield checker. A checker in the outfield wastes somewhere between 4 and 5 pips depending on its position. So we add the point it's on first to our total of 15 already. Here that checker is on 14 point so 15+14=29. From there we still need to add 4-5 pips of wastage for that specific checker so our total is between 33 and 34 as the final epc of the top player.
The point of last take is equal to the number of rolls to go minus three. Here we have a 4 roll position as mentioned earlier to the point of last take would be 1 effective pip or 4-3=1. We said our final epcs were 30/34 so it's a clear pass. Too easy isn't it?
Now it's time for you to play around with that outfield checker and see how good the formula holds up. Remember the bottom player's epc will always remain 30 (30.4 if you want to be exact/check with a bot) and the point of last take will remain the same at ~1 effective pip or a count for the player considering taking of 31.4. The key positions to inspect will of course be the outfield checker on the 13pt and then on the 12pt. It's amazing to me how useful this stuff is and how I ever survived without it. I would make wild ass guesses in these positions where now I'm confident that the worst thing that will ever go wrong is for me to make the tiniest of errors.
on roll, cube action?
is Player 1 
