Seeking node counter tool

[mjlef]

Now that I am a bit more active in programming again, I want to tune my program's move ordering. What I would like is a tool to run a big bunch of positions through, search them to fixed depth, then count up the total nodes searched on my UCI program. Is there a nice simple tool to do this?

Mark

[Bill Rogers]

Most programs just include a variable within their program to count each new node as the program generates it. Mine only counts legal move nodes
Bill

[MattieShoes]

I don't think there's an external tool specifically for this. It should be relatively simple to write a function that flips through positions and keeps a running tally of nodes though.

One possible alternative is to keep a tally of how many moves were searched before the current one whenever score>alpha. If you're not searching an enormous number of positions, this could be more useful, maybe...

[Bill Rogers]

Isn't that exactly what I said?
Bill

[Gian-Carlo Pascutto]

No, what you said had nothing to do with his question.

He wants a program that takes his engine, runs it through positions, and adds the nodecounts.

He obviously already has a node counter inside the program.

[Jim Ablett]

Now that I am a bit more active in programming again, I want to tune my program's move ordering. What I would like is a tool to run a big bunch of positions through, search them to fixed depth, then count up the total nodes searched on my UCI program. Is there a nice simple tool to do this?

Mark

Hi Mark,

Thomas Mayer's modified Epd2Wb by Bruce Moreland can also handle Uci engines and does
pretty much all you've asked for. It shows node count for each solution as it goes along. Total
nodes should be easy to add to it. (C src included)

Download:
http://www.mediafire.com/?nmynigzmg4y

Jim.

[mjlef]

Thanks!

[sje]

What you need are two tools: a driver that runs your program with specified search parameters on a problem set, and a statistical analysis program that looks at the results.

If your program can generate EPD and the statistics part of the solution is EPD driven, then half of the task is solved. This approach has the benefit of using common code and of not being specific to the UCI protocol.

Symbolic has an interactive command interface that substitutes for the xboard (or UCI) interface for most development work. This interface includes batch file command support, the automated running of EPD test files, and the scoring of the EPD test result files. This is the approach I recommend to others although it requires a bit more work up front.