Winboard and UCI - The Big debate Part II

Home |Articles | Links | Guestbook| Chat | Bookmark|
Monitor page
for changes

   it's private   

by ChangeDetection

Note this is part 2 of a article discussing the pros and cons of Winboard/UCI. You can find Part 1 .

However according to Shapiro, Carl and Varian, Hal R. (1999): Information Rules: A Strategic Guide to the Network Economy,factors other than pure technical merit decide the victor in a standards war. The following factors (besides technical quality) is listed as capable of affecting the failure or success of standards.

  1. First mover advantage
  2. Control over installed base of users
  3. Ability to innovate
  4. Brand recognition
  5. Complements (probably irrelevant)

We shall keep these factors in mind :0)

Ability to innovate

As mentioned above, some people think that the strict UCI protocol where the GUI controls much of the interface is restricting for Chess engine authors who want to try something different for say pondering. Another area, where Winboard protocol appears to be ahead is that it has a nice system based around the protover and feature system that allows new features to be added while remaining backward compatible.

In additional a mailing list discussing such changes to the protocol is available . This is in contrast to UCI of course where no such discussions to improve UCI has as yet occurred.

On the other hand, UCI is very new (barely over 1 years old), compared to Winboard's nearly 10 years, so it's natural that the aging protocol is in need of changes.

In fact, the Winboard protocol is pretty much a ad-hoc protocol ,with it's origins as a interface to support GNUchess and no other engine. This need to support older engines makes it's a little difficult to make changes, compared to UCI which is a brand new interface built from the scratch.

Whether Winboard or UCI will dominate will depend on how quickly and easily the two protocols evolve to handle the weakness (some of them detailed above) of their respective protocol. Currently , it's unclear if either protocol is likely and easier to innovate faster.

First mover advantage, Control over installed base of users

In terms of the first mover advantage, Winboard clearly holds the edge. It has being around since 1994 first as a informal standard and then slowly as it caught on (see my timeline for more details) it became the dominant standard in 98-00. It's closest competitors at the time Auto232 by Chris Donningers simply couldn't compete because Winboard protocol was much easier to use, more robust especially on one computer (which most users had).

UCI however is a latercomer introduced at the end of 2000. Initially it did not catch on for several reasons. Technically it wasn't all that superior to Winboard (for reasons already mentioned above), and Winboard had already established itself as a dominant standard. Winboard's popularity and a free Online client also meant that it had a large installed base of users. UCI on the other hand was handicapped because it had a very small base of users. Only one interface, Shredder 5 supported UCI and Chess Assistant 6 (which supported Winboard standard as well anyway). Both were commercial. Naturally, UCI didn't really catch on, and few engine writers bothered to support UCI, primarily because it didn't make much sense to waste resources on supporting a protocol that was not widely used because only 2 interfaces supported them (Chess Assistant and Shredder 5). Interface writers also had little incentive to support UCI because there were few UCI engines. A classic chicken and egg problem.

As such after a while it seemed that UCI was a standard that was dead before it could take off.While a few Winboard engines (Yace and Pepito) supported the standard on top of Winboard, this number was a mere drop in the bucket compared to the number of Winboard engines, and few commercial engines besides SOS,Shredder seemed to join in either.(The top commercials all belonged to the closed Chessbase protocol of course)

Gambitsoft's attempt to sell "Winboard Edition II" a collection of WB/UCI engines , was the only addition to the UCI family in May 2001.(It's perhaps significant that the packaged was named Winboard Edition II, which confused many buyers who thought Winboard II was a improved winboard. And the fact that Gandalf, the strongest of the bunch on the CD was available only as UCI (because the earlier CD released already had the Winboard version), only angered many buyers who assumed Gandalf would be available as a Winboard version.)While a second package of UCI engines was planned, it was eventually cancelled.

As such at the end of Dec 2001 , almost one year after UCI was announced in Nov 2000, the number of UCI engines stood at less than 10, compared to the Winboard engines (both commercial and free) which has surplus the 100 engine mark in May 2001.And the 150 engine mark about a year later

All this changed with one stroke, in early 2002 as a result of 2 events. Firstly, Chessbase released a new patch for Fritz 7 (which had being released at the end of 2001), which allowed it to support UCI. This occurrence was not totally unexpected, as Stefan Meyer-Kahlen had joined the Chessbase team, and Shredder 6 had earlier released Shredder 6 in both a UCI interface as well as Chessbase. Another interesting signal was that Chessbase dropped providing the much criticised Winboard adaptor from their site for download in Dec 2001. In retrospect, they were preparing to jump ship to support UCI.

Chessbase's support gave UCI standard a new lease of life. Being the most popular Chess software company for serious chessplayers (Chessmaster8000 actually sells the most number of copies, probably more then all the sales of others combined ), it has a large base of experienced users. This alone meant that UCI standard was suddenly built into a interface with thousands of users. Suddenly, engine writers had a incentive to support UCI. Besides the fact that more users would be using UCI and the fact that the UCI engines appeared to work better then Winboard engines in Chessbase interfaces (but it's still buggy see this for a very serious problem ) meant that engine writers could test their engines at full strength against the World Class Chessbase stable of Engines (Fritz,Junior,Shredder etc). They could of course request to create a Chessbase engine, but in most cases, it would involve giving up their source which many did not want to do.

Still as popular as Fritz is, it's still a commercial interface, and not everyone was/is willing to buying a commercial interface just to use UCI. Arena by Martin however solved this problem neatly.

Arena which went into beta testing in Jan 2002, is a free interface that supports both UCI and Winboard protocols. As such, even engine authors who didn't want to spend cash on buying a commercial UCI interface had a UCI interface to test against. However it was not exactly a UCI equal avant of Winboard/Xboard for 2 reasons. Firstly, it was not under the GNU license. And secondly, so far it works only in Winboard, and Chess engine authors who work on Unix/Linux are out of luck. The second reason, is significant as currently, there is still no UCI interface that works on non-Windows platform.

The rise in UCI - between Jan 2002 to June 2002

How fast has UCI risen? Here is some data.

Period As at Nov/Dec 2001 As at June 2002 Increase %
UCI interfaces 3 8 166%
UCI engines 11 24 118%

Period As at Nov/Dec 2001 Jan 2001 to June 2002 Increase %
UCI Engines (free) 4 18 350%
UCI engines (commercial) 7 6* -
Total 11 24 118%

* The number of commercial UCI engines fell by one, because the once commercial Lampchop 10.xx was released for free.

In 6 months, the number of UCI engines/Interfaces has more than doubled. The number of UCI interfaces can be attributed to other commercial companies (Chesspartner,Rebel's Gandalf GUI, Chess Academy 7 as well as the free Arena) playing catch up with the market leader Chessbase to ensure that they don't fall behind. Only Chessmaster is still lagging behind (it's quite unlikely the next version Chessmaster 9000 will support UCI) , but this is to be consistent for a mass market product given that Chessmaster was also one of the last major database/playing program to implement Winboard support.

The increase in number of UCI engines in percentage growth is impressive, however in the same period the number of Winboard engines grew by about 40! Since there are far more free Chess engines that commercials, it is expected that much of the growth in UCI engines must come from existing new Winboard engines adopting the standard, or brand like engines adopting the standard.

Breakdown of free UCI engines

Type Engines Total
Part of Winboard Edition III Nejmet,Anmon,Pharaon,Dragon 4
Available for free before as Winboard engines Yace,Pepito,Little Goliath,Engimax,Sjeng,Queen,Amyan 7
Newly released as both Winboard/UCI Monarch,Hagrid,Aristarch,Big Lion,Abrok 5
Released only as UCI for free SOS for Arena 1
Total - 17*

*The table excludes Siboney which is currently not available and possibly a clone of Pepito and Lampchop 10.xx which was formerly commercial ( in Winboard edition II it was available as both Winboard and UCI as well as a weaker free public winboard version) but now is free.

The first category consists of Winboard engines that began to support UCI in preparation to be sold as a commercial package of UCI engines as far back as May 2001.Technically, their release for free later (because the project was cancelled), is not in any way due to the "Fritz 7 and Arena effect" but as it happens they are also Winboard engines with UCI support tagged on.

It's interesting to note that of the remainder released, every single one of them except one (SOS for Arena - a freeware version of the commercial SOS), supports Winboard as well.Given that there are currently over 100 Winboard engines and the similarities between Winboard and UCI it's no surprise that a large number of UCI engines are actually Winboard engines with secondary UCI support.

The second group which consists of long time released Winboard versions is likely to increase, (eg Terra,Quark may join the group soon), but it's interesting to see how many of the actively worked on projects will jump onto the UCI band wagon as well. It's hard to be sure, but it's unlikely any of them will drop Winboard support for UCI only, although the sole exception appears to be Little Goliath.

The third group consists of existing Winboard engines that was not released publicly untill recently. Aristarch is a example where the author decided to add UCI support as well before releasing it, so it's primarily a Winboard engine as well.

However for UCI to become a threat to Winboard, it would mean that programmers would design their new engines to run only in UCI.

But Even Monarch and Hagrid which are almost brand new Engines built from the scratch support both protocols. Monarch however in it's initial stages had only partial Winboard support.

Future Prospects

It's hard to make any conclusions from the minimal data above.But it seems clear that because of the similarity and ease in supporting both, most Chess engine authors will begin to support both protocols. Given the current lead in Winboard engines, I expect, Winboard protocol to continue to hold the upper hand, but UCI may catch up, when some of the existing actively developed Winboard engines start to add UCI support.

Still,I doubt if the Winboard protocol will die off anytime soon and it's likely that both standards will co-exist. The Winboard protocol could be in danger though if we see new engine authors supporting only UCI, or existing Chess Winboard engines authors switching to support only UCI. As yet, there are still no signs of such things happening yet.

Aaron Tay

Update Dec 2002

The total number of engines supporing UCI has now increased to slightly over 30. The number of UCI only (with Winboard support) engines now stand at 6. 2 are mater solvers (Chest and Mater),  The others are SOS Arena,Delphimax,Silkchess,Goliath (formerly Winboard only).