Association of British Scrabble Players

Analysing Games Using Maven


Maven is a computer program written by Brian Sheppard. It is no longer commercially available, the rights having been bought by Hasbro to produce their official CD-ROM which uses the North America word list. For legal reasons, this website cannot suggest any means of obtaining a copy of Maven which plays to the international word list. Members of the uk-scrabble forum may be able to help.

Maven comes with a powerful tool for analysing Scrabble positions, its Simulator. The following guide to analysing Scrabble games has been written by former ABSP Chairman Pete Finley.


Introduction

Several people have asked me to explain how to use Maven to analyse their games. Until now my reply has always been that they should read the help files on ANNOTATING GAMES and SIMULATOR and then get back to me if they had any specific queries. However, I remember how much trouble I had in figuring out how to use Maven, so I have decided to prepare what I hope is a simple step-by-step guide to the process. My own computer skills are very limited, so this should be straightforward enough for anyone to follow.

I have assumed that you have managed to get Maven installed and working. This is merely an explanation of how to analyse games.

I submitted my first draft of this guide to a group of leading players - David Webb, Andrew Fisher, Phil Appleby, Ed Martin and Stewart Holden - for their comments. I am very grateful for numerous suggestions that have been incorporated into this revised draft.

The information you need

To analyse a game you have already played, you need the details of all moves played by both you and your opponent and you need to know your own racks for each move. If you want to analyse your opponent's moves too, then you'll also need to know what their racks were. Most people, however, are content to analyse their own moves.

Late game and endgame analysers

If you read the relevant Maven help files, I recommend you ignore anything they say about using the late game and endgame analysers. They almost invariably cause my computer to crash and I know other players have had the same problem. There is one procedure you should follow when you reach the late stages of the game, but I will explain that later.

Equity loss - what it means

In using Maven to analyse a game, you are assessing the mistakes you made compared to an ideal game where you played the best move on each turn. The equity points you lose can be viewed as extra points of spread you could have gained over the course of the game. Adding your total equity loss to your spread (which will be either positive or negative depending on whether you won or lost the game) will give you an estimate of what the result could have been. For example, if you lost by 60 points but lost 80 points of equity, then it is theoretically possible that you could have won that game by 20 points with perfect play.

This is of course merely a theoretical assessment. As soon as you played one move that was not the ideal move, then the game changed. The board changed and the letter sequence changed, so the subsequent moves you played might not have been possible in the ideal game. Nevertheless, it is interesting to look at equity loss as a rough guide to how much you could have changed the result of the game.

ANALYSING A GAME

Getting started

Open up Maven and click on FILE - NEW to start a new game.

If no word appears on the board diagram, you can skip the rest of this paragraph. However, if Maven plays a word onto the screen, click STUFF - ALTER POSITION and a smaller board will appear over the top of the screen. Click on the first letter of the word played, then drag the cursor in the direction of the rest of the word. By pressing the space bar key you can delete the word from the board. Alternatively, you can place the cursor to the right of the word on the board and use the backspace key to delete it. You should now have an empty board diagram and you can start analysing your own game.

If you played the first move, you can skip the rest of this paragraph. If your opponent played the first move, type it in place on the board. You can change the direction of type from horizontal to vertical by moving the cursor in the direction you want to go. It's important that you get the positioning of the word right, because a wrong positioning will affect the result of subsequent move simulations and you will have wasted your time.

Now you need to enter your own rack in the box underneath the small board diagram, labelled YOUR RACK. Obviously you should first delete the rack that is there already. When you've entered your rack, click OK and the small board diagram will disappear. You will see that your rack is now displayed to the right of the large board diagram.

Using the kibitzer

Now click STUFF - KIBITZ and a list of possible moves will appear underneath your rack. Check that the move you actually played is amongst them, making sure that it is being played in the position that you played it. A grid reference is given after each move, but you can easily check where the word is being played by double-clicking on it. The word will then be put in place on the board. To clear it from the board, simply click underneath the kibitzer's list.

If your move is in the list, then you can skip the rest of this paragraph. If the word you played is not on the list of moves, you can add it to the list by dragging the tiles from your rack into the right position on the board, or simply by entering the cursor at the correct starting point and typing the word. In the process, you may add other words to the move list that you do not want to be included in the simulation. If so, you can remove them from the list by clicking and dragging on the relevant words in the list to highlight them, then clicking on EDIT - CLEAR.

If you need to add a change of tiles to the list, you can do this by dragging the appropriate tiles into the bag marked PASS. Tiles have to be dragged one at a time, so you may add extra changes to the list that you do not want to be included in the simulation. You can remove these in the way described at the end of the previous paragraph.

Preparing, running and stopping the simulation

When the move list (sometimes called the kibitz list) is as you want it, click STUFF - SIMULATE. A box titled Simulation Options will appear. You don't need to tick any of the boxes down the left hand side. The only box you need to consider is the one at the top that indicates how many moves ahead you want the program to simulate. I usually leave it at one. Some people, however, prefer to set it at two. You can set it for any number you wish, but the higher the number you choose, the longer the simulation will take. It is best to set the look ahead option to two moves if any of the options under review involves a change of tiles.

When you have got the Simulation Options box the way you want it, click OK and the simulation will start. You'll see a number counter above the move list that tells you how many iterations have been done. You can let this run to any number you wish. I always use 1000 iterations for analysis of my tournament and club games, though I sometimes use just 300 iterations for quick analysis of other games and positions. The more iterations you run, the more accurate the simulation results will be. However, many players are happy with 300-500 iterations, looking one or two moves ahead.

The only way to stop a simulation is by pressing ESCAPE on your keyboard.

Simulations do take a while to complete, especially if you are running 1000 or more iterations. Some players like to speed up the process by running a preliminary simulation of perhaps 100 or 200 iterations and then deleting those moves that are clearly out of the running. If you do this, make sure that you leave your own move in! You can delete moves, as described above, by highlighting them in the kibitzer and using EDIT - CLEAR. When you have removed the no-hopers, run another simulation to however many iterations you are using for your full analysis. You should note, however, that the new simulation will start afresh. It will not continue from where you stopped the preliminary simulation. The fewer moves you leave in the simulation the faster it will run. However, beware of taking too many moves out after the preliminary simulation, as things can sometimes change considerably between say 100 and 1000 iterations.

It is not always necessary to run the simulation to the same number of iterations. If your chosen move is well ahead at some point before that, then it is reasonable to stop the simulation early, as the gap is highly unlikely to be closed completely. For example, though I use 1000 iterations generally, I will stop the simulation if my move is 10 points ahead after 300 iterations or 5 points ahead after 500. Other players may well use different guidelines.

Calculating equity loss

When you have stopped the simulation, if the move you played is at the top of the list, then congratulate yourself because you played the best move and lost no equity on that turn. But if your move is not at the top of the list, subtract the number it achieved in the simulation result from the number that the top move achieved. The result of that subtraction is your equity loss for that move. For example, if the best move had a simulation result of 36.55 and your move had a simulation result of 32.44, then in theory you lost 4.11 points of equity on that move. For practical purposes, many players simply round each move loss to the nearest whole number.

Moving on

It is important at this point to enter the move that you did play by double clicking on the word in the kibitz list. That will add your move to the board diagram. Then click STUFF - ALTER POSITION, add your opponent's next move to the board, enter your new rack into the YOUR RACK box, click OK and you are ready to repeat the analysis procedure for your next move. Click STUFF - KIBITZ and follow the simulation procedure through again. After each of your moves, calculate your equity loss for that move and add it to your cumulative total.

The late game procedure

When you reach the point when there are 12 tiles or fewer left unseen (5 in the bag and 7 on your opponent's rack) there is a procedure you should follow to maintain the accuracy of the simulation. When you click SIMULATE and the Simulation Options box comes up, change the look ahead setting from one move to 100 moves. This should be done for any moves played when there are between 8 and 12 tiles left unseen.

Calculating endgame equity loss

When there are no tiles left in the bag, you cannot simulate the moves any more. But the process of calculating equity loss is not over. This is how you calculate the equity loss for moves from that point. Enter your rack as normal and click Kibitz. Put on the board whatever the kibitzer says is the best move (Maven's endgame play is its strongest feature) and then follow the game through to the end from there, playing whatever Maven says is the best move for both you and your opponent. NB Ignore what you both actually played if it was different - just follow the optimum moves to the end of the game. Work out how many you would have won or lost by at the end of that sequence of ideal moves.

Then erase that theoretical endgame from the board, input the move you actually played, if it was different, and follow the optimising procedure through to the end again. Once again, you are only concerned with the optimum moves. Work out how many you would have won or lost by and compare it with the figure you got the first time. The difference is your equity loss for that move.

If you played any moves after that then you repeat the process for those moves too. It can get very annoying because you can mess up the endgame two or three times and take losses each time.

Totalling up your equity loss

When you have added the equity loss for your final move to your cumulative total you will have your equity loss for the game. In theory, the lower the figure, the better you played the game. You will find that your total losses for different games will vary enormously. At the time of writing this, March 2005, my best effort is a total equity loss of 12. My worst is a terrible 321. Ed Martin has similar best/worst figures of 7 and 303. In tournament games I average about 95 points of equity loss. For comparison, Stewart Holden averages around 100, Ed Martin around 97, David Webb around 70 and Andrew Fisher around 55, while recent WSC and BEST finals have seen average losses of 50 or so. We all have slight variations in our simming procedures, but the impact of these variations on our average loss over many games is negligible.

It is worth noting that the players named above are all SOWPODS players. Equity losses tend to be lower for players who play to the American word list. This is because there are fewer words in that list and it is easier to end up with a blocked board, which restricts options.

Learning from your equity loss

You can learn much more from analysing your games than just how many equity points you lose overall. It is important to understand why the best move IS the best move. By dividing your equity loss into different types of mistake you can pinpoint which areas of your game need most improvement.

I confess to being a little lazy in this respect and trusting to general awareness of the type of mistakes I make most often. Some players are much more rigorous in their approach. Ed Martin, for example, divides his equity loss into three types:

  • bad evaluation of moves (where he spotted the move but rejected it)
  • overlooking the best move (where he knew the word but didn't spot the move)
  • not knowing the word for the move

For the last category, Ed further categorizes the words involved by their length. He then pays special attention to the categories in which he makes most mistakes.

A FEW SPECIAL POINTS

Simulating challenged phonies

The best way to calculate equity loss for a challenged phony is to add an exchange of each letter on your rack to the kibitz list. When you have run the simulation, count the middle scoring tile exchange as your move and take the equity loss for that move.

If there are fewer than seven tiles in the bag you cannot follow the procedure above as you are not allowed to change tiles in that situation. There is no really satisfactory procedure in these circumstances - my best suggestion is to penalise challenged phonies at this point by the actual points score for the move.

Stewart Holden bypasses all of the above by simply counting every phony played as an equity loss of 50pts, to cover the average score for a move together with the lost opportunity for rack balance and the knowledge of one's rack given to the opponent. However this is less accurate than the above methods.

Unchallenged phonies

I believe that if you play a phony and your opponent does not challenge it, you should still penalise yourself, using the method outlined above, because, under the free challenge rule at least, a phony is a mistake and it should have been punished. Many players agree with this, but not all. David Webb, for example, says: "An unchallenged phony is a potential not an actual source of equity loss. I like to maximise the correlation between equity loss and points loss." David prefers to treat unchallenged phonies as a valid word simmed in the normal way.

The procedure for adding a phony to the kibitz list is:

  • enter the word in the Alter Position board and click yes when Maven asks if it is good
  • go back to the Alter Position board and delete the phony
  • now click Kibitz, retype the phony on the board and it will appear in the kibitz window.

Moves where you know your opponent's rack

If your opponent's last move was a phony bonus, which you challenged off, then you know their rack. In these circumstances you have to be very careful in calculating equity loss for your subsequent move. If your opponent still has a playable bonus then any move that does not block that bonus may well be a bad move. You have to use your judgement here and decide whether what the simulation says is the best move really is the best move.

Late game moves that may be "winninger" than Maven's suggestion

This is so difficult to assess that many people, including myself, don't even try to make any judgements in this respect and we count any equity loss Maven says we should have, even if we don't agree with it. Some people do choose to occasionally override Maven's opinion, but even those who do this don't do it very often. There is a program called Simwin, developed by US player John O'Laughlin, that can calculate the "winningness" of a move, but for some reason it won't run on my computer. If you want to try it, it's available at http://john.morphism.org/simwin

CONCLUSION

These notes have been primarily concerned with the practicalities of analysing games with Maven. It is worth reading through the help files, which will give you an insight into the theory behind the analysis. The section on HOW TO PLAY WELL will give you particularly useful advice. However, beware the section on Tile Values (which should really be entitled Rack Leave Values), as the values given are for the American word list. The SOWPODS values are:



A +1			N +0.5
B -2			O -0.5
C -0.5			P -1
D +0.5			Q -9
E +2			R +1
F -2			S +8
G -3			T  0
H +1.5			U -4
I -1			V -6
J -2			W -3
K  0			X +2.5
L +0.5			Y -0.5
M +0.5			Z +5

Blank = +25


When analysing your games, it is important to remember that calculating your equity loss is not the main purpose of the exercise. Average equity loss over a large number of games is an excellent indicator of how well you play the game and is therefore worth knowing, not least because an appreciation of the amount of equity you lose can have the beneficial effect of reducing complaints about luck. However, the main point of analysis is to help you become a better player. To achieve this, you need to understand why you are losing equity and why one move is better than another, so that you can address your shortcomings, avoid repeating your mistakes in future games and so reduce your average equity loss.

I hope this guide helps you improve your game with Maven. Using Maven has certainly improved mine.

Pete Finley




Data Protection Act: In joining the ABSP I accept that personal information relating to Scrabble, including my rating and my photograph, may appear on the ABSP website and in ABSP publications. I accept that the ABSP will not be responsible for use or misuse of personal information.

SCRABBLE® is a registered trademark. All intellectual property rights in and to the game are owned in the U.S.A. by Hasbro Inc., in Canada by Hasbro Canada Inc. and throughout the rest of the world by J.W. Spear & Sons Ltd. of Maidenhead SL6 4UB, England, a subsidiary of Mattel Inc. Mattel and Spear are not affiliated with Hasbro or Hasbro Canada.

ABSP © 2013

Contact Us