What do you think are the top 3 engines in order of strength

That usually means one engine is being less thorough on the way down than the other (aggressive pruning).

Different engines might not be reporting the same thing as the ply though. For example, Houdini reports the deepest ply reached, while Rybka reports the average (or the minimum, or something else other than deepest). Given this, it's quite possible for Rybka to actually have searched deeper but report a lower ply than Houdini, and I wouldn't be surprised if similar differences hold true for other engines.

Chess engine writers have to make trade-offs: how fast a position is evaluated compared with how 'well' it is evaluated. A low knode/s suggests the engine is doing better evaluations instead of quicker approximations. A human GM evaluates maybe a few nodes per second but still manages to play decent chess.

Same goes with search depth. All chess engines cut away branches they guess will not lead anywhere. The more you cut, the higher depth you achieve -- but with risk missing out on good variants that got cut off. Or sometimes look deeper into the crucial lines. Is depth the deepest branch looked at, the average, or the depth at which all branches have been evaluated (and only cut off 'exact' and not cut off speculatively)?

Also how the depth is reported varies. For instance Rybka 3 reports depth-4 (it says 9, but means 13) -- since a depth of 4 is minimum for alpha-beta pruning to kick in (and maybe the Rybka 3 engine cannot work with ab disabled?).

Rybka gives false search depth figures to hide its roots is from the Fruit Program

Having roots from other programs shouldn't be much of a big deal, when it still leads to improvement.

Houdart also said that he took the basics from some open-source-engine.

Taking something that is good and making it better is much more effective than starting over from scratch.

Is that the fifty third thread of the kind?

Rybka has been banned from competing, for being a copy of Fruit. While copying as such can be ok, but Fruit was released under the GPL -- which would make Rybka GPL too (and hence illegal to sell, and allowed to be distributed for free).

Houdini is allegedly a clone of Robbolito, which is an open source engine, based on reverse engineering Rybka 3.

So they are all copying each other, and guess improvement is good. It also leads to a stagnation of ideas. There are currently so many variants of Fruit/Robbolito/Ivanhoe clone engines out, so it is hard to be top 200 without:

• being another Fruit clone
• using very similar algorithms (since the competitions have rules for what parameter settings you have to use, and if your engine is too different to even have the same parameters, you are not allowed to compete)

But guess I am ranting off-topic now.

I messed up with the numbers? Was it fifty four then?

That's impressive!