Links
BETA Testers about REBEL10
written by Claudio Bollini
Introduction
Since my last review (many months ago) some important events have happened in the Computer Chess world. The tense rivalry between the few best programs has now generalized into a Big War of many contenders.
The history is widely known: For about eight years, Lang’s programs inside Mephisto dedicated computers reigned unbeaten. Then the first master-level program for PCs: Mchess, appeared and had its days of glory too; next, the Chess Machine (with Schroder’s and de Köning’s engines) and Wchess got their own ephemeral zenith.
But soon it turned evident that the age of the absolute supremacy of a single program was over. In the following years it took place a sort of stabilization, while a bunch of 4 or 5 clearly excelled over the rest, surpassing each other by a minimal edge.
Now we are assisting to another turning point. Quite often some low profile program, or perhaps a rising star from amateur arena, jumps right into the elite, making room by its own, and pushing for the first place in the SSDF rating. The days of an unarguable leader were already over, but nowadays even the stable club of top 5 seemed to have been abolished, receding at an ever-growing tight and little crowd.
Let’s put aside the polemic around the former lower-strengthen Fritz, and if it deserves the pole position or not. Newcomers to hilltop like Shredder, Junior, Nimzo, and a bit below Gandalf, and Kallisto, have made their meritorious way through. And there are many others exercising hard, while they wait in the lane their opportunity to cross their swords.
To face this situation, the traditional and well-respected Mchess, Genius, Hiarcs and Rebel have had to stress their resources to keep seizing the market. Once a program reaches the peak, subsequent developments leading another 100 Elo points boost becomes no longer likely. So the programmers and their teams have now a hard time fine-tuning the engines, carefully adding controlled amount of fresh knowledge, implementing subtle modifications to searching algorithms.
And they feverishly keep trying, wishing to earn that beloved 80-100 Elo jump. Most of times, however, they are not able to announce any spectacular strength improvement. Therefore, they reinvest an important part of the precious time until the deadline, in aesthetic retouches to GUI, better management of Databases, hugger books, and may be some cutting-edge sellable feature; and resigned themselves to release the engine with a slightly (sometimes, imperceptible) improvement.
As I said, Rebel belongs without doubt to the crème de la crème, and it didn’t make an exception to this crude marketing strategy of release-or-die. In my previous Rebel review, I was concerned with this lacking of a net difference between former and newer engines. Indeed, better GUIs and more powerful data resources are great, but what about the core of every chess program: the engine itself?
I disagree with those who state that any further engine development is almost futile. Yes, I know that current commercial programs can crush the 99% of us. But following the same idea, we can also object as unworthy the improvements over database programs like ChessBase, because it has already got more features than most of tournament players would ever need. Chess programs have still a long way to go, pursuing better positional understanding i.e. of those quiet positions hiding long-term goals.
For instance, it would be a valuable project the building of an interactive tutorial capable of teaching GM-level strategic concepts.
Well, let’s finally arrive to what really matters: I believe that Ed Schröder has achieved a largely expected, substantial, qualitative advance with his Rebel 10.
New Graphical User Interface
Rebel’s team has made a valuable effort in this aspect. Although it is still a DOS program, they have done a serious try to approach it to Windows style. Although not fully resizable, Rebel interface allows now a wide range of options to get a user-friendly layout, like selectable board sizes; turning on/off and movement for the different windows; and settings of the length and number of various items.
Rebel 10 can be configured to be quite informative about its internal thoughts. Now it is possible to watch at the same time the scrolling PVs and the teacher windows. The teacher option is a very useful complement to the numeric evaluation, since it allows peeking deeper into Rebel’s evaluation function composition. I prefer this natural language over the supposed more thorough auto-annotation of Fritz 5, or even worse, the one of Chess Master 6000.
The teacher window shows two group of items, one based on individual status of pieces (pinned and hang pieces, isolated and passed pawns), and the other referring to general aspects. The first is clear enough; but the other classification, although sharp and useful, isn’t so easy to interpret. Here we have four items: king safety, your mobility, passed pawn chances and pawn structure.
Since the comments are the averaged result of individual estimations for white and black sides, there are not mayor problems with pawn evaluations, but the first two remains almost always confusing. Besides, are these comments based on the root or on the terminal position? As the previous version, Rebel 10 on-line help derives the user to readme file for further explanations, but this file doesn’t mention anything about the issue.
Databases functions
There are two database modes. One, more limited, to manage pgn files, with basic functions but translatable to Rebel format. The another one is for Rebel databases themselves, and it has been vastly improved.
As for Rebel 9, the database screen includes a pair of boards representing for the game selected the last move until out-of-book, and the terminal position. The terminal board could be evaluated within the database screen.
Besides, it has incorporated to usual data fields (player name, ECO, Elo, result, date, etc.) a new index freely definable by the user, to make flexible searches and sorting.
Another considerable progress over previous version: Boolean searches are now allowed, via searchmask option into database screen. Clicking on any of three rows of options: OR, AND and NOT, you can easily add or discard any of these items.
You have also the possibility of defining a searching range (MORE THAN and LESS THAN). Finally, there are four little boards on the bottom, allowing simple definitions on moves, positions, pattern and material, respectively.
In short, the database working environment is much better organize and flexible, and some new resources lead this new version to a high standard.
The new Encyclopedia of Chess (EOC)
The standard CD was released with a carefully selected 300K database of games (or more impressively talking, a 16 million of unique positions) and stored in a quick access format, as the base for the EOC feature.
Rebel Company offers an additional EOC CD with 50 million of positions. (They state in their Web Page that this is the biggest chess tree available till now). With the use EOC knowledge option, you can set it either to your own information, in a consult window much like the Fritz 5 tree, or as a Rebel book.
The manual does not recommend you to use EOC as an opening book for Rebel itself, but more as an option to study opening theory.
The EXTRA option
This is the informal aspect of Rebel, just for funny purposes: simultaneous games, energy levels, or blindfold play. I enjoy them, and don’t think that they attempt at all against the seriousness of the program.
However, I sincerely believe that the promoted ultra-secret option for blitz chess is a bit useless and disappointing. The feature adds or subtracts time laps for certain type of predefined moves, like captures or pawn moves. I tried it, and I cannot figure out how can it be helpful for improving chessplayer strength. It’s only adrenaline, but without real chess.
The Chess Engine
I spent a lot of PC overnight processing time, running several sets of positions with the new epd autotest facility. Now I got many statistical tables, and I have noticed a few more on CCC forum. It’s clear for me that these are only tests, and that they cannot totally measure a strength enhancement for a living game.
The victory against Anand is also impressive, but still it isn’t enough to get the big picture. However, both tests and match games served at least to confirm my feelings.
I mean, you will only realize how much better this version is when you confront it in a plain game. Rebel 10 is not only tactically sharper and faster, but mainly it is more positionally talented.
Its famous anti-GM feature (please, do not turn it off; just leave it as smart and let Rebel to manage it), offers an original and surprising aggressiveness, not usual for the traditional Rebel personality. Rebel 10 will try to push you, if feasible, into semi-open or even purely open positions, and once there it will apply a sound strategic criterion.
The positional sacs have become even more probable than for its elder brothers, and its well-known human style, even more human than ever.
The amazing fact is that this metamorphosis from active to aggressive behavior has not been done in detriment of its marvelously well-balanced style. I believe that Ed got here a real benchmark.
The combination option activates a special algorithm for tactical purposes, but it is only available for single positions. Therefore, the obvious drawback is that you couldn’t use it for a whole game or for living games. And for the latter I would prefer to take a glance to the pretty, genuine Rebel 10 engine, capable of providing perhaps the most accurate evaluation available in a chess program.
By the way, to get a closer Elo estimation of actual Rebel 10 through epd set of tests, you have to set combination option off. Although the combination mode is faster, getting thus the right move earlier, this algorithm makes the engine strategically blinder at a given fixed time.
Additional endgame knowledge has also been added, but according my tests, you will notice the betterment of the specific know-how at this stage, mainly in slow time controls.
Conclusion
- Rebel 10 is an optimal choice for a wide range of chessplayers, from serious amateurs to grandmaster level tournament participants.
- Despite some minor drawbacks, its data management capabilities have reached a word-class performance, and its GUI layout could be considered now truly user-friendly.
- The EOC feature provides a valuable help for opening preparation, and there is an extensive amount of information available to follow deeper the analyses of Rebel’s brain.
- However, the most important earning has to be conceded to the engine itself. Its pure playing strength has had a noticeable boost, perhaps the most important since old Gideon was released.
- But for another point of view, there is a new style arising, still more enterprising and human-like, that can be properly described as a successfully balanced trade off between tactical sharpness and strategic understanding. A wonderful alliance.
- It is difficult to predict if Rebel 10 will reach the first place in the Swedish List. But, anyway, it isn’t so important; they are only computers playing each other, while you and me are human beings. In my opinion, there is a fact much more meaningful: Rebel 10 has surely achieved both the strength and the good sense of a strong human grandmaster.
Ordering information
In general REBEL10 (new) costs $59.95, REBEL10 (upgrade) costs $29.95. For specific valuta check the Rebel price list.
The below two companies ship REBEL software all over the world, allow all possible payments like VISA/MASTER etc., are known for good service and fast delivery. You can email Gambit Soft (Germany) or ICD Your Move (New York) by clicking on the companies logo for remaining questions or to enter your order.
Order from Gambit Soft (Germany) attention Bert Seifriz.
Order from ICD Your Move (USA) attention Steve Schwartz.
If you want to order from your local dealer then check out the REBEL dealer list.
REBEL dealer list (by phone)
REBEL dealer list (by email)