Can anyone explain Houdini's Tactical Mode?

Can anyone explain Houdini's Tactical Mode (version 3 and later) in full detail? Houdini's website gives a very basic idea of what it's supposed to do, but does not provide much in the way of specifics. Does that mode scale down when lowering engine strength to make for a different style of play at lower simulated ELO settings, allowing variety between positional and tactical play, or is it purely only useful for analytic use? In otherwords, does the engine play sufficiently different with Tactical Mode turned on than when it's turned off at lower strength settings?

Thanks,

Joe

Houdart might be the only person who can definitively answer your questions. If I were the Houdini programmer, I would want the tactical mode to only be active at full strength, but that's just me.

Try sending him an e-mail; Maybe he'll answer it.

P.S. - You could probably test this out yourself. Just have two clones play each other in a number of games. Both engines would be set to a low elo level. They would be identical except that one would have the tactical mode activated.

If, after a large number of games, the results show that the engine strengths are statistically equal, then you would have your answer.

Unfortunately, I don't have the engine to test. I'm currently using an older version of Houdini, prior to the introduction of Tactical Mode. That specific feature intrigued me and would be incentive for me to purchase Houdini 5 if it works the way I hope it does. If it only works properly with the engine at full strength, then it wouldnt meet my needs as I typically use Houdini as a sparring partner with the strength scaled down to approx 1600 ELO.

I'm not sure what you mean. I haven't played any games on chess.com, only locally on Fritz and Aquarium against a few choice engines. I'm just asking for specifics about the Tactical Mode feature in Houdini.

I used to play competitively about 15 years ago, and my rating back then was about 1850. I'm recently getting back into chess again, but mostly as a hobby, not competitively. I dug out my DGT board, upgraded it to USB, and am using it for sparring against AI. I'm playing against Houdini 1.5e with an adjusted strength parameter of 13 (about 1400 simulated ELO) with an average W/L ratio of approx 60%, so I've relapsed and am not able to play at 1850 anymore. That's about the only metric I can provide for now. I'm not looking for the highest rated engine (which right now would be Stockfish 8 and Komodo 11). I'm just curious about tactical mode.

I had this question myself.  Here is the answer as I now understand after some research.

If by "how it works" you mean "what is it doing at the engine level," a thread in tactical mode prunes "bad" lines much less aggressively.  Once in a great while, one of those lines turns out not to be bad after all.  The benefit is it will not miss tactics within its search depth.  The cost is it will have a much reduced search depth.

If you mean "how does it allocate threads between the two modes," I can give you two explanations, one of which is correct.  I am not sure which one, but I THINK it is the second one.  This is very annoying and unfortunately typical of chessbase which tends to have bad documentation exactly along these lines.

The two possibilities are:

1)  When you go into advanced settings you can set a number, which will be the number of threads that will be in tactical mode.  This is what "most people" think.  I believe it is wrong, but I can't prove it.  Anyway, it is one of the possibilities.

2)  When you set the number, it is the proportion of threads that will be in tactical mode, not the number.  Crazy.  Only a computer programmer would think this is reasonable.  So, select 1 and you get the fraction 1/1, which means 100% of threads are tactical mode.  2 gives 1/2 which means half of the threads are tactical mode.  4 gives 1/4, so every fourth thread in tactical mode.  Etc.  This is the answer which I believe is correct.

It's easy to find "information" about Houdini tactical mode if you consider "transforms it to the greatest tactical analysis tool ever" to be information.  If you consider that to be marketing baloney and want actual nuts and bolts information, it's hard to find.  However, explanation 2 above comes from what I *think* is correct documentation after a long search.  On the other hand, maybe someone made it up and now I'm spreading it.

My solution is to set Houdini for 4 threads and tactical mode 2.  That way it's 2 tactical and 2 regular threads both by number and by proportion.

Looks like you are correct...

http://www.cruxis.com/chess/manual/index.html

That's exactly what I was looking for!  Thanks!

I was looking for information on the same topic and my answer's about a year late, but here's what it says in the Houdini 6 Online Manual (http://www.cruxis.com/chess/manual/index.html):

Tactical Mode (checkbox)

Activate tactical mode threads,

Default is 0, min 0, max 8 (Standard) or 128 (Pro)

0 = no tactical mode

1 = all threads use tactical mode

2 = every second thread uses tactical mode (50% of the threads will be in tactical mode)

4 = every fourth thread uses tactical mode (25% of the threads will be tactical)

etc.

When Tactical Mode is active, Houdini modifies its search process and uses some clever tricks to focus on tactical solutions in the root position. In Tactical Mode Houdini becomes a very skilled tactical problem solver.

Below some results for two popular tactical test suites with Houdini running at 20 seconds per position with 2 threads on a Core i7-5820K @ 3.9GHz using 512 MB hash.

Arasan 19 Tactical Suite, comprising 200 medium to hard positions

- Houdini 4 (normal mode) solves 123 positions, total solution time 34:59 minutes.

- Houdini 5 (normal mode) solves 124 positions, total solution time 35:30 minutes.

- Houdini 6 (normal mode) solves 154 positions, total solution time 25:34 minutes.

- Houdini 4 Tactical Mode solves 158 positions, total solution time 24:31 minutes.

- Houdini 5 Tactical Mode solves 168 positions, total solution time 20:57 minutes.

- Houdini 6 Tactical Mode solves 172 positions, total solution time 16:17 minutes.

Sedat test suite of 100 positions

- Houdini 3 (normal mode) solves 50 positions, total solution time 23:05 minutes.

- Houdini 4 (normal mode) solves 42 positions, total solution time 24:17 minutes.

- Houdini 5 (normal mode) solves 57 positions, total solution time 22:30 minutes.

- Houdini 6 (normal mode) solves 62 positions, total solution time 19:31 minutes.

- Houdini 3 Tactical Mode solves 72 positions, total solution time 15:47 minutes.

- Houdini 4 Tactical Mode solves 72 positions, total solution time 15:37 minutes.

- Houdini 5 Tactical Mode solves 74 positions, total solution time 14:21 minutes.

- Houdini 6 Tactical Mode solves 76 positions, total solution time 12:42 minutes.

Note that the focus on tactical moves will reduce the playing strength of the engine, the Tactical Mode is significantly weaker than the normal playing mode.

Using Tactical Mode value 2 (meaning that half the threads are tactical, assuming you're running at least 2 threads) will cost about 20 Elo.

One can compare the Tactical Mode to what a human player does when he/she tries to solve a test position in a chess magazine or website. He/she will scan for tactical, forcing moves, threats, sacrifices etc. That way a lot more tactical solutions can be found than what the player would actually see and decide to play over the board, during a normal game. The "this is a test position" message changes the search strategy.

Houdini's Tactical Mode works exactly the same, very human-like: it will spend a lot more time looking for tactical moves, threats etc. In most positions this doesn't provide any benefit - most of the time there is no tactical solution available and the increased focus on tactics is just wasted. But in test positions (selected because there IS a tactical solution) it works extremely well.