Fritz ratings vs.

6/8/2006 – We receive lots of questions about Fritz 9's rated game mode – and about just exactly how does that Elo rating system work? Steve Lopez tackles these questions and tells you why your Fritz rating isn't neccearily identical to your over the board rating in the latest ChessBase Workshop.

ChessBase 14 Download ChessBase 14 Download

Everyone uses ChessBase, from the World Champion to the amateur next door. Start your personal success story with ChessBase 14 and enjoy your chess even more!


Along with the ChessBase 14 program you can access the Live Database of 8 million games, and receive three months of free ChesssBase Account Premium membership and all of our online apps! Have a look today!

More...

We're going to cover two topics in this ChessBase Workshop: how to set up and play a game in Fritz' rated game mode, and how this rating stacks up against a rating you'll earn in OTB (over the board, i.e. "face to face") and/or online play.

First things first. Let's look at how to play a rated game in Fritz.

You need to realize straight away that rated game mode is a much more restrictive mode of play than any of the others which Fritz offers. The point of rated game mode is to simulate as closely as possible the conditions that you'll face in a face to face rated tournament or match game against a human opponent. This means that you don't get takebacks, you don't get to "peek" at what Fritz is thinking via the Engine analysis pane, you can't get help from an engine "kibitzer", you can't pause the game, and you can't stop it to save it and come back to it later. Once you start a rated game in Fritz, you have to play it out to a conclusion in that sitting: win, lose, or draw. If you bail early on a rated game, it's scored as a loss for you. That's why you should always select a time control that you can complete in a single session.

That last point is an important one -- and I've argued it several times in chess message boards. If you can pause or stop a rated game, the whole point of the exercise becomes moot. It's not a true test of your skill if you can jack up your rating by pausing or stopping the game to consult with a book, a chess engine, or a strong human player. Besides, you'd only be cheating yourself.

Setting up a rated game is pretty easy. Fire up Fritz, click "Play Fritz", hit F3 and select the engine you want to play against, and set your time controls (you'll do this by selecting File/Blitz or Long game, and setting the time parameters). Then go back to the File menu, select "New", and then "Rated game" from the submenu:

A fair little bit of this dialogue is self-explanatory. The upper righthand box contains your name, the rating you've earned so far, your won/lost/drawn record, your overall score, and how much "money" you have. This "virtual money" thing and the "doubling" checkbox deal with a form of faux gambling [*gasp*]. You're essentially betting play money on the outcome of the game. If the "doubling" box is checked, you can ask Fritz at any time during the game if it wants to double the dollar amount of the wager on the game's outcome. If Fritz refuses, you win on the spot (the game and the wager). Likewise, Fritz can ask you the same thing and, if you refuse, it's the same as resigning. In my opinion, it's essentially a meaningless feature geared toward the hardcore gamblers amongst us, and essentially turns chess into a poker-like "bluff" game. But it's there if you want it.

"Use clock" is a very un-tournamentlike option, but puts an interesting twist on things -- neither player can lose by running out of time. But this also means that both you and Fritz have an unlimited amount of time to ponder a move, and this can make for a very loooooooong game in some cases. If you plan to play a timed game, please be sure to set the clock before entering into rated game mode.

Finally you get a rating slider which lets you select the approximate rating for Fritz. This range of ratings is a function of your computer hardware as well as the engine you've selected (older engines tend to be somewhat weaker than the newer ones). If you want Fritz to play against you full-bore, make sure you've checked the "Unleashed" box -- you'll see the approximate rating for this setting after the word "Unleashed" in the dialogue.

When you're ready to go, click "OK" and the game will commence. Fritz will alternate colors from game to game (if you had White in your last rated game, you'll have Black in this one). When the game is finished, you can go to File/New/Rated game and see a new rating which has been computed based on your latest result:

You won't get a rating until you've completed at least twenty rated games. And you might see Fritz' rating range on the slider drop a bit if you're not doing well against the program.

Now you know how to do it. That's part one. The second part of this article deals with your Fritz rating and how it translates to a "real life" rating you might get by playing in rated chess events.

In short, it doesn't.

Now if I was as big a jerk as I pretend to be, I'd just type "Until next week, have fun" and let you hang. But my comment merits a bit of explanation, so we'll press onward.

There's no direct correlation between any two chess ratings you might have which were obtained by playing different forms of chess (OTB, correspondence, and online) and/or within two different pools of players. A rating is simply a gauge of your results within a particular closed pool of players. It's not an absolute -- it changes each time you play -- and it doesn't "transfer" from one organization/website/program to another.

You see people ask the question online all the time: "I have a rating of xxxx at [online site] -- how would this compare to a USCF rating?" or, more recently, "I have a rating of xxxx against Fritz -- how would this compare to a [national federation/FIDE] rating?"

I'm going to present now a piece I wrote on this subject several years ago for my personal chess website. I'm doing this for two reasons: 1) the article has been unavailable for several years now, ever since I folded that site and 2) I'm a really, really lazy guy who doesn't like to have to do the same thing twice. Although the article deals specifically with how an online rating "compares" with an OTB rating, some of the same reasoning applies to your games against Fritz compared to an OTB rating.

This piece also explains how the Elo rating system works (and I get asked about that a lot, too). So, from 1999, here's my old article about the chess rating system:


If you're a regular reader of the Usenet chess newsgroups or other chess "bulletin boards" on the Internet, I'm sure you've seen the question asked dozens of times: "How does my online chess rating compare to an official USCF rating?" The answer is that it doesn't. There are a few reasons why you can't compare the two, but before we get to them we need to look at how the rating system works.

The current chess rating system used by the USCF, FIDE, and most online chess servers is called the Elo rating system (named for its creator, mathematician Arpad Elo). The Elo system is a convenient way to compare the skill levels of two players to see which player is the most likely to win. It greatly rewards upset victories while minimizing the risk for lower-rated players losing to much more highly-rated players.

In the pure form of the Elo system, a new player has a rating of zero. The result of his first game will determine his initial provisional rating to be used as a number for averaging purposes. A win will give the new player a first provisional rating 400 points higher than that of his opponent. A draw will give him a rating equal to that of his opponent. A loss will give him an initial rating 400 points lower than his opponent.

Let's look at this in action. In his first rated game in a USCF, a new player (who we'll call Fred Patzer) faces an opponent rated 1600. Should Fred win, his first provisional rating will be 2000. If he draws it will be 1600, and if he loses it will be 1200. Fred loses his first game and winds up with an initial rating of 1200.

Next we'll look at how this process works in Fred's first four game weekend tournament. He lost to a 1600-rated player in his first game (as we've seen). In his next game, he wins against a player rated 1326. This gives him a score of 1726 to be averaged in with his existing rating of 1200 as follows:

  • 1200 + 1726 = 2926
  • 2926/2 (the number of games he's played so far) = 1463

So after two games, Fred Patzer's provisional rating is 1463.

In round three, Fred is paired against a player rated 1544 and manages a draw. The way his provisional rating is now figured works like this -- his current provisional score of 1463 is used twice (signifying his average result over the first two games), while his new factor of 1544 is added in to the total. Since he's played three games so far the total is divided by three to get his new provisional rating:

  • 1463 +1463 +1544 = 4770
  • 4470/3 = 1490

After three games, Fred's provisional rating is now 1490.

In round four, Fred plays a 956-rated player with disastrous results. He loses and gets a factor of 556 to be merged in with his current provisional rating as follows:

  • 1490 +1490 +1490 + 556 = 5026
  • 5026/4 =1256.5 rounded up to 1257

So after his first USCF tournament, Fred's rating is now 1257.

This procedure will be used until a player completes his provisional rating period. The length of this period varies from league to league. The USCF requires 20 games as a provisional period for over the board play and 25 games for postal play. Leagues in other countries may have different requirements.

Going into Fred's twentieth over the board game, he has a provisional rating of 1472. He faces a player rated 1320 in his last game and draws. His first official rating is calculated as follows:

  • (1472 x 19) + 1320 = 29288
  • 29288/20 = 1464.4 rounded down to 1464

Fred Patzer's first official rating is 1464.

Once a player finishes his provisional period and gets an official rating, his subsequent results are calculated differently. It's a bit more complex than the provisional system, but isn't too hard to grasp. Here's the basic formula:

  • Rn = Ro + .04(ED) +/- 16

This means that a player's new rating [Rn] is calculated by starting with his old rating [Ro], adding 4% of the difference between his rating and his opponent's rating, and adding or subracting 16 points. You lose points when you lose a game and gain points when you win one. For example, Fred's current rating is 1464. He beats a player rated 1586. His new rating is calulated as follows:

  • 1464 + 5 + 16 = 1485

Fred has gained 21 rating points to attain a new rating of 1485. Meanwhile, his opponent has lost the same number of points (21) to drop his rating down to 1565.

Had Fred lost this game, his rating would have been figured this way:

  • 1464 + 5 - 16 = 1453
Fred's rating would be 1453 while his opponent's would have been 1597.

In the event of a draw, the plus or minus factor of 16 would have been dropped from the equation. Fred's rating would then be 1469 and his opponent's would be 1581.

There are various tweaks to the rating system. No player can gain or lose more than 32 points or less than 1 point in a game, regardless of the rating differential. Players rated between 2100-2399 gain or lose fewer points than normal (12 is used as a base factor instead of 16), while players rated 2400 and above use 8 as a base factor.

This is basically how the USCF's rating system works. The basic principles apply to all Elo-based rating systems for over the board, postal, and online chess. However, there are a couple of reasons why there is not a direct one-to-one correspondence between USCF over the board ratings and the ratings you get from online servers.

A minor reason is the base rating assigned to a new player. You are supposed to have a rating of zero when starting out as a new player. However, many online servers give a base rating of 1500 or 1550 to a new player. In most cases this is not critical. But it does become a factor in rare cases when two new (unrated) players face each other. If it's a server in which a new player is assigned a 1550 rating, a loss for one player gives him an averaging factor of 1150, while the winner has one of 1950. In both cases this is entirely too high when compared to USCF ratings in a similar circumstance. In such cases, the USCF assumes a 1200 rating for both players, providing provisional values of 800 and 1600 respectively in the case of a decisive result. The online server has already skewed the results +350 points, resulting in rating "inflation" compared to USCF ratings.

The major reason you can't draw a comparison between an online rating and a USCF rating, however, is actually very simple and doesn't involve major mathematical calculations.

When you go to a USCF over the board tournament, who you'll play is determined by the pool of participating players and your game results, according to a set of rigid guidelines. For example, our friend Fred Patzer (now rated 1485) goes to a chess tournament at the local community center. Let's say that most of the players there are masters (rated 2200+). Fred will almost certainly face a master in the first round. He has the possibility of gaining 32 points if he wins and will only drop one rating point in the event of a loss.

However, let's assume that all of the players who show up are rated lower than 1100. Fred finds himself in a very weak tournament and has a lot to lose: he'll only gain 4 points total in this four round event (assuming he wins all of his games) and can lose as many as 128 points on the day if he loses every game.

Admittedly, these are extreme examples. But the point is that in an over the board tournament, you don't get to pick who you'll play (this depends entirely on who's there as well as on your round-to-round results), you don't get to pick when you'll play (the starting times for the rounds are predetermined), you don't get to pick how many games you'll play (unless you opt to drop out of the tournament early), you don't get to choose your color each game (in theory, you should have an equal number of games as White and Black, but in online play you can play all your games as White if you so choose), you don't get to choose the time controls under which you'll play (these, too, are predetermined), you don't have the option to ask for takebacks (touch-move is rigidly enforced), and you don't have the option to try to distract your opponent by chatting with him during the game (talking is prohibited, unless it's to adjust a piece, offer a draw, announce check, or resign).

Compare this to online chess, where anything goes. You can pick your opponents at will, select your own time controls, play any time of the day or night, play as many or as few games as you desire, and merrily chat up your opponent as you play the game.

By far, the most significant of these factors is the ability to choose one's opponents. In online chess, you need never play a lower-rated opponent if you don't want to, thereby minimizing your risk. By playing nothing but opponents rated higher than yourself you can gain points in the long haul. If you do nothing but play opponents rated 400+ points higher than your current rating, you have a chance to gain 32 points each game for a win while only dropping a single point should you lose. Even if you win just one of every twenty-five games, your rating will improve, even if only by a few points. It's almost completely risk free.

Some players specialize in draws; exchanging pieces like madmen to suck the life out of a position. Such players are bound to increase their rating by playing only highly-rated players -- a draw in such a case is worth 14 points to the lower-rated player while a loss still results in the loss of just a single point.

Other players jack up their initial ratings by playing games only against opponents rated 2200 or higher during their provisional period. Here's a sample game I played on a chess server. My rating (as White) in this game was 1654. My opponent (Black) was rated a whopping 2711. Take a look at this game and tell me if you think my opponent was a real-life grandmaster:

1.e4 e6 2.d4 d5 3.e5 Nc6 4.Qg4 f5 5.exf6 gxf6 6.Qh5+ Ke7 7.Nf3 Nh6 8.Ng5 Kd6

1-0

This game was played at a Fischer time control of 25/10. There is no excuse for a 2711-rated player to handle a game this way. My opponent was probably rated about 700 in real-life over the board chess and had inflated his rating during his provisional period by only playing opponents rated 2500 and up.

Then there are players who specialize in fast time controls. They learn a limited (often bizarre) opening repertoire and hammer their opponents with the same opening setup in blitz game after blitz game. They become experts in the positions that arise from this oddball opening they've practiced and can move almost without thinking, while their opponents are left trying to find their way as the precious seconds dwindle away.

Beyond this, there's the often raised issue of cheating. It's nearly impossible to cheat in a well-run over the board tournment, while the methods for cheating in online chess are myriad and well-documented.

It's easy to see that there is no basis for direct comparison between online ratings and official USCF over the board ratings. A clever online player can manipulate his provisional rating by carefully selecting his competition and thereafter maintain an artificially high rating for an extended period by using the techniques outlined above. The only way to actually gauge your true rating is by playing officially rated over the board chess.


I suppose you're now going to ask the question: "If my Fritz rating isn't the same as a real OTB chess rating, what good is it?"

It's good for nothing if all you care about is the number.

I don't play OTB rated chess anymore and haven't for a decade -- I have jobs (yes, plural) and kids and responsibilities and other hobbies and a social life and there's just no damn time for it anymore. A pity, really. So I content myself with playing rated online correspondence games. While there's certainly a difference in the manner in which they're played, there's absolutely no difference in my approach to them.

A rated game against another player (whether carbon or silicon) is nothing but a test. When you're sitting down to play a rated game, you're testing yourself -- it's your skill and ability against those of your opponent. Sink or swim, do or die, win or lose. Sure, you're testing him but -- more importantly -- you're testing yourself.

And the only danged thing a chess rating represents is a gauge of how well you're doing. If you're succeeding, if you're learning, your rating will go up. If you're not, your rating drops. A rating isn't an absolute -- it's fluid. It'll go up or down depending on how you're doing within that pool of players -- or, in the case of Fritz, that particular set of chess engines (since you can use more than one engine in rated game mode).

All a rating consists of is a measure of your performance -- a guide to how well you're learning, a graph of how you're doing. It's not an end in and of itself. When I play a game, I'm not worried about its affect on my rating; I'm just trying to pass the test by beating my opponent and, in a lot of cases, beating myself. Will I rise to the challenge or will I crack? All the number tells me is whether or not I'm growing as a player. It's just a gauge.

And your Fritz rating is just a gauge of how well you're doing in your games against Fritz. That's pretty useful for people who can't get out to rated tournaments, find postal correspondence a financial expense, and don't have an online connection (or a good, dependable one).

I could go on, but by now I'm either preaching to the choir or it's all falling on deaf ears. If all you care about is a rating, a number, whether it's your Fritz rating compared to your USCF rating, or your online rating compared to your correspondence rating, or your OTB rating compared to Jimmy's down the street, you've missed the whole point of this beautiful game. If so, that's pretty sad.

Until next week, have fun!

You can e-mail me with your comments on ChessBase Workshop. All responses will be read, and sending an e-mail to this address grants us permission to use it in a future column. No tech support questions, please.


© 1999, 2006, Steven A. Lopez. All rights reserved.


Topics: f9
Discussion and Feedback Join the public discussion or submit your feedback to the editors


Discuss

Rules for reader comments

 
 

Not registered yet? Register