What's Vasik Rajlich doing today?

[Rebel]

I wasn't talking about Rybka, just highlighting your condemning and patronizing remarks, the red above.

Again, eh? I have seen ZERO cases of a complete rewrite. Throw out the old code and start from scratch with nothing. That is a dumb way of developing software and that is not a "condescending remark" it is a simple statement of fact. If someone DID rewrite something N times from scratch, they are really very poor programmers. Once, perhaps, to change language, data structures, board representation and everything else. But more than once? What a waste...

So by your logic Fabien and Marcel are dumb programmers, bravo Bob.

Many realities out there but only the Hyatt reality counts.

Here is another reality, mine. Somewhere in the 90's I wrote an experimental version from scratch. The basic ideas;

1. Incremental update of the evaluation function (gain speed).
2. Don't generate all moves at once, just one at the time when you need one (gain speed).

Made no sense to use the regular Rebel code because it was fully ASM, such a fundamental different approach one does in C first and when it has a future you convert the critical parts to ASM for the final speedup.

All the reasons in the world to start from scratch.

I have YET to see where anyone has rewritten from SCRATCH. Not once. So no "bravo, bob" but rather "bravo Ed for wanting to divert the topic rather than paying attention to what has been claimed." "from scratch" means ZERO code (none, nada, etc) retained from previous version. Doesn't happen very often. Certainly not more than once for a single programmer.

Try again, failure here.

Again, the premise has been "more than once". I'll give anyone one complete rewrite after they learned what they did wrong the first time...

I just gave you an example, but feel not to believe me

And I may suspect you know what it means to write an "incremental upate" engine, right? You can't use anything any longer, not even the move generator. So you start from scratch.

And BTW, for clearness sake, you started this discussion with the word "ENGINE". That excludes own interfaces, protocols like WB or UCI.

[bob]

I wasn't talking about Rybka, just highlighting your condemning and patronizing remarks, the red above.

Again, eh? I have seen ZERO cases of a complete rewrite. Throw out the old code and start from scratch with nothing. That is a dumb way of developing software and that is not a "condescending remark" it is a simple statement of fact. If someone DID rewrite something N times from scratch, they are really very poor programmers. Once, perhaps, to change language, data structures, board representation and everything else. But more than once? What a waste...

So by your logic Fabien and Marcel are dumb programmers, bravo Bob.

Many realities out there but only the Hyatt reality counts.

Here is another reality, mine. Somewhere in the 90's I wrote an experimental version from scratch. The basic ideas;

1. Incremental update of the evaluation function (gain speed).
2. Don't generate all moves at once, just one at the time when you need one (gain speed).

Made no sense to use the regular Rebel code because it was fully ASM, such a fundamental different approach one does in C first and when it has a future you convert the critical parts to ASM for the final speedup.

All the reasons in the world to start from scratch.

And that's the exact case that Bob gave where writing from scratch makes sense.

Eh, I don't think you understand the magnitude of change that involves moving to an "incremental update" type of chess engine and (likely) with it the need (or self-preservation hehe) to do it in C first.

This is all well and good. But what does it have to do with Rybka / Fruit? There is no asm anywhere to be found. The issue here is software development for a chess engine, which is no different than software development for any other complex system. I don't need to give a list, but operating systems do NOT start from scratch for each major or minor version. Compilers do not start from scratch. Check out gcc 5.0 which is a significant new change. Compare it to one of the later 4.x versions (I think I am using 4.8.4 but am not sure). No complete replacement. Lots of code reuse. Which is basic software engineering.

When Mark compared the 2.3.2a evaluation to 1.0 beta, he found differences which he has mentioned (imbalance table, etc). But he ALSO found lots of similarities, which is to be expected. Crafty is now about 20 years old. On version 25.0 at the moment. MAJOR changes along the way, from doing the old incremental bit board stuff as done by Slate in chess 4.x through rotated bit boards as I did them, thru Pradu's magic move generation stuff. But there have never been more lines of new code than of old code.

The point for going into this diversion is beyond me. Why argue about what "rewrite" means? It is irrelevant in the context of Fruit and Rybka (or in the case of Glaurung 1, Glaurung2 and Stockfish, because anyone can look at the source to see that G2 is not 100% new code. It has been significantly modified, but parts of G1 are laying around everywhere. Tord is not a dummy programmer, so throwing away good code just to start over is not something he (or any other good programmer) would do. You might mix and mash it a lot to improve readability, modifiability or maintainability, but not throw everything out, unless it was just lousy code.

In the case of Rybka, we do have one known example of this, namely when Val threw out the pre 1.0 version and started over with Fruit. But apparently even there he did not throw everything from 1.6.1 out, he kept the crafty move generation stuff and I don't know what (if anything else) as I didn't look and don't care, since the case has been resolved and is now a part of past history.

In short, none of this has anything to do with Vas and Fruit / Rybka. Never did. Never will. It is a "hope" some hold that perhaps R4 is clean via the complete rewrite from scratch idea. It is a very "faint" hope since 2.3 and whatever did not appear to be sloppily written, to justify throwing everything out and starting over... And even that is moot for this case since no mention of Rybka 3/4 matters since we know that 2.3.2b participated in the ICGA event, possibly 2.3.2a as well.

[bob]

I wasn't talking about Rybka, just highlighting your condemning and patronizing remarks, the red above.

Again, eh? I have seen ZERO cases of a complete rewrite. Throw out the old code and start from scratch with nothing. That is a dumb way of developing software and that is not a "condescending remark" it is a simple statement of fact. If someone DID rewrite something N times from scratch, they are really very poor programmers. Once, perhaps, to change language, data structures, board representation and everything else. But more than once? What a waste...

So by your logic Fabien and Marcel are dumb programmers, bravo Bob.

Many realities out there but only the Hyatt reality counts.

Here is another reality, mine. Somewhere in the 90's I wrote an experimental version from scratch. The basic ideas;

1. Incremental update of the evaluation function (gain speed).
2. Don't generate all moves at once, just one at the time when you need one (gain speed).

Made no sense to use the regular Rebel code because it was fully ASM, such a fundamental different approach one does in C first and when it has a future you convert the critical parts to ASM for the final speedup.

All the reasons in the world to start from scratch.

I have YET to see where anyone has rewritten from SCRATCH. Not once. So no "bravo, bob" but rather "bravo Ed for wanting to divert the topic rather than paying attention to what has been claimed." "from scratch" means ZERO code (none, nada, etc) retained from previous version. Doesn't happen very often. Certainly not more than once for a single programmer.

Try again, failure here.

Again, the premise has been "more than once". I'll give anyone one complete rewrite after they learned what they did wrong the first time...

I just gave you an example, but feel not to believe me

And I may suspect you know what it means to write an "incremental upate" engine, right? You can't use anything any longer, not even the move generator. So you start from scratch.

And BTW, for clearness sake, you started this discussion with the word "ENGINE". That excludes own interfaces, protocols like WB or UCI.

Why would that be? Cray Blitz was a heavily incremental update program. It updated pawn structure counters (isolated, doubled, passed pawns) and such, move generation information and such. I removed the incremental stuff in the 80's for simplicity and I didn't rewrite everything. Perhaps it depends on how the code is designed and written, but for me it wasn't an issue.

Engine includes whatever plays chess. From book moves, to hashing, to evaluation, to searching, to making/unmaking moves, to detecting check, generating moves, you name it. My basic move generator hasn't changed in a LONG time, even though the methodology for generating moves has. And for clarity, I did talk about "COMPLETE rewrite from scratch" which would be necessary to lose parts of fruit in later rybka versions. Others jumped in with the bogus stuff about glaurung and such, based on half-assed interpretations of what an author wrote / meant and then bending that interpretation to try to make a point that can't be made realistically.

[bob]

What does it mean to rewrite code? One does not start from scratch as there are experiences from previous work. So even if I sit down t redo a task done before I will do some things as before. Why? Because I want to focus on those points that caused me to rewrite at all.

Starting a chess program "from scratch" means to me one wants to get rid of doing things in a way just because they were done so. But I can't imagine to start everything new because change only because of change is a waste of work.

In this threat it is just arguing for words as there are many definitions for "starting from scratch". This arguing is bad for computer chess but the repetitions from all sides within the last few years show that there won't be an end...

A semantic war is what it means.

In general use, "rewrite" does not mean "throw out the old and start from scratch to write the new" any more than "rebuild means throw out the old and start from all new parts." You will hear people talk about writing and re-writing a Ph.D. dissertation several times before it is finished.

Another term is "write from scratch." or, if you wish "rewrite from scratch", were one would mean starting from nothing, the latter would mean throwing everything existing out you have previously written and again starting from nothing."

experience doesn't really factor in, although you make a point there, in that once you do something one way, chances are you will do it the same way the next time you write such code from scratch.

[Rebel]

I wasn't talking about Rybka, just highlighting your condemning and patronizing remarks, the red above.

Again, eh? I have seen ZERO cases of a complete rewrite. Throw out the old code and start from scratch with nothing. That is a dumb way of developing software and that is not a "condescending remark" it is a simple statement of fact. If someone DID rewrite something N times from scratch, they are really very poor programmers. Once, perhaps, to change language, data structures, board representation and everything else. But more than once? What a waste...

So by your logic Fabien and Marcel are dumb programmers, bravo Bob.

Many realities out there but only the Hyatt reality counts.

Here is another reality, mine. Somewhere in the 90's I wrote an experimental version from scratch. The basic ideas;

1. Incremental update of the evaluation function (gain speed).
2. Don't generate all moves at once, just one at the time when you need one (gain speed).

Made no sense to use the regular Rebel code because it was fully ASM, such a fundamental different approach one does in C first and when it has a future you convert the critical parts to ASM for the final speedup.

All the reasons in the world to start from scratch.

And that's the exact case that Bob gave where writing from scratch makes sense.

Eh, I don't think you understand the magnitude of change that involves moving to an "incremental update" type of chess engine and (likely) with it the need (or self-preservation hehe) to do it in C first.

This is all well and good.

Thanks for the admission.

But what does it have to do with Rybka / Fruit?

Nothing, I already said that, see the green above.

You started this "from scratch" discussion making disparaging remarks on Marcel's multiple rewrites of Rookie from scratch. Nothing to do with Rybka.

[Rebel]

I wasn't talking about Rybka, just highlighting your condemning and patronizing remarks, the red above.

Again, eh? I have seen ZERO cases of a complete rewrite. Throw out the old code and start from scratch with nothing. That is a dumb way of developing software and that is not a "condescending remark" it is a simple statement of fact. If someone DID rewrite something N times from scratch, they are really very poor programmers. Once, perhaps, to change language, data structures, board representation and everything else. But more than once? What a waste...

So by your logic Fabien and Marcel are dumb programmers, bravo Bob.

Many realities out there but only the Hyatt reality counts.

Here is another reality, mine. Somewhere in the 90's I wrote an experimental version from scratch. The basic ideas;

1. Incremental update of the evaluation function (gain speed).
2. Don't generate all moves at once, just one at the time when you need one (gain speed).

Made no sense to use the regular Rebel code because it was fully ASM, such a fundamental different approach one does in C first and when it has a future you convert the critical parts to ASM for the final speedup.

All the reasons in the world to start from scratch.

I have YET to see where anyone has rewritten from SCRATCH. Not once. So no "bravo, bob" but rather "bravo Ed for wanting to divert the topic rather than paying attention to what has been claimed." "from scratch" means ZERO code (none, nada, etc) retained from previous version. Doesn't happen very often. Certainly not more than once for a single programmer.

Try again, failure here.

Again, the premise has been "more than once". I'll give anyone one complete rewrite after they learned what they did wrong the first time...

I just gave you an example, but feel not to believe me

And I may suspect you know what it means to write an "incremental upate" engine, right? You can't use anything any longer, not even the move generator. So you start from scratch.

And BTW, for clearness sake, you started this discussion with the word "ENGINE". That excludes own interfaces, protocols like WB or UCI.

Why would that be? Cray Blitz was a heavily incremental update program. It updated pawn structure counters (isolated, doubled, passed pawns) and such, move generation information and such. I removed the incremental stuff in the 80's for simplicity and I didn't rewrite everything. Perhaps it depends on how the code is designed and written, but for me it wasn't an issue.

Engine includes whatever plays chess. From book moves, to hashing, to evaluation, to searching, to making/unmaking moves, to detecting check, generating moves, you name it. My basic move generator hasn't changed in a LONG time, even though the methodology for generating moves has. And for clarity, I did talk about "COMPLETE rewrite from scratch" which would be necessary to lose parts of fruit in later rybka versions. Others jumped in with the bogus stuff about glaurung and such, based on half-assed interpretations of what an author wrote / meant and then bending that interpretation to try to make a point that can't be made realistically.

I will elaborate a bit since we are having a misunderstanding what I mean with "incremental update" of EVAL. In a normal chess engine (depending on the size of EVAL the engine is in EVAL for 50-90% of the cpu time.

In an "incremental update" engine you don't evaluate positions but moves. You modify the eval score with the changes a move causes. And it's super fast. I did this in my first chess engine (1980) in BASIC, it could even (and was more or less obliged since the absence of a QS) evaluate if a checking move was a checkmate.

I guess you will now understand the implications of such an approach.

[Darkmoon]

I wasn't talking about Rybka, just highlighting your condemning and patronizing remarks, the red above.

Again, eh? I have seen ZERO cases of a complete rewrite. Throw out the old code and start from scratch with nothing. That is a dumb way of developing software and that is not a "condescending remark" it is a simple statement of fact. If someone DID rewrite something N times from scratch, they are really very poor programmers. Once, perhaps, to change language, data structures, board representation and everything else. But more than once? What a waste...

So by your logic Fabien and Marcel are dumb programmers, bravo Bob.

Many realities out there but only the Hyatt reality counts.

Here is another reality, mine. Somewhere in the 90's I wrote an experimental version from scratch. The basic ideas;

1. Incremental update of the evaluation function (gain speed).
2. Don't generate all moves at once, just one at the time when you need one (gain speed).

Made no sense to use the regular Rebel code because it was fully ASM, such a fundamental different approach one does in C first and when it has a future you convert the critical parts to ASM for the final speedup.

All the reasons in the world to start from scratch.

I have YET to see where anyone has rewritten from SCRATCH. Not once. So no "bravo, bob" but rather "bravo Ed for wanting to divert the topic rather than paying attention to what has been claimed." "from scratch" means ZERO code (none, nada, etc) retained from previous version. Doesn't happen very often. Certainly not more than once for a single programmer.

Try again, failure here.

Again, the premise has been "more than once". I'll give anyone one complete rewrite after they learned what they did wrong the first time...

I just gave you an example, but feel not to believe me

And I may suspect you know what it means to write an "incremental upate" engine, right? You can't use anything any longer, not even the move generator. So you start from scratch.

And BTW, for clearness sake, you started this discussion with the word "ENGINE". That excludes own interfaces, protocols like WB or UCI.

Why would that be? Cray Blitz was a heavily incremental update program. It updated pawn structure counters (isolated, doubled, passed pawns) and such, move generation information and such. I removed the incremental stuff in the 80's for simplicity and I didn't rewrite everything. Perhaps it depends on how the code is designed and written, but for me it wasn't an issue.

Engine includes whatever plays chess. From book moves, to hashing, to evaluation, to searching, to making/unmaking moves, to detecting check, generating moves, you name it. My basic move generator hasn't changed in a LONG time, even though the methodology for generating moves has. And for clarity, I did talk about "COMPLETE rewrite from scratch" which would be necessary to lose parts of fruit in later rybka versions. Others jumped in with the bogus stuff about glaurung and such, based on half-assed interpretations of what an author wrote / meant and then bending that interpretation to try to make a point that can't be made realistically.

I will elaborate a bit since we are having a misunderstanding what I mean with "incremental update" of EVAL. In a normal chess engine (depending on the size of EVAL the engine is in EVAL for 50-90% of the cpu time.

In an "incremental update" engine you don't evaluate positions but moves. You modify the eval score with the changes a move causes. And it's super fast. I did this in my first chess engine (1980) in BASIC, it could even (and was more or less obliged since the absence of a QS) evaluate if a checking move was a checkmate.

I guess you will now understand the implications of such an approach.

Who the hell cares about any of this? If Rajlich came back on to the scene tomorrow and presented a Rybka UCI updated version, that he stated was competitive with the now top three chess engines-I'd be online with Pay Pal.

And none of Dr. Hyatt's repetitious guilty tripping about Vas being a code copying cheat would make the slightest difference in my decision to do otherwise.

[Darkmoon]

I wasn't talking about Rybka, just highlighting your condemning and patronizing remarks, the red above.

Again, eh? I have seen ZERO cases of a complete rewrite. Throw out the old code and start from scratch with nothing. That is a dumb way of developing software and that is not a "condescending remark" it is a simple statement of fact. If someone DID rewrite something N times from scratch, they are really very poor programmers. Once, perhaps, to change language, data structures, board representation and everything else. But more than once? What a waste...

So by your logic Fabien and Marcel are dumb programmers, bravo Bob.

Many realities out there but only the Hyatt reality counts.

Here is another reality, mine. Somewhere in the 90's I wrote an experimental version from scratch. The basic ideas;

1. Incremental update of the evaluation function (gain speed).
2. Don't generate all moves at once, just one at the time when you need one (gain speed).

Made no sense to use the regular Rebel code because it was fully ASM, such a fundamental different approach one does in C first and when it has a future you convert the critical parts to ASM for the final speedup.

All the reasons in the world to start from scratch.

I have YET to see where anyone has rewritten from SCRATCH. Not once. So no "bravo, bob" but rather "bravo Ed for wanting to divert the topic rather than paying attention to what has been claimed." "from scratch" means ZERO code (none, nada, etc) retained from previous version. Doesn't happen very often. Certainly not more than once for a single programmer.

Try again, failure here.

Again, the premise has been "more than once". I'll give anyone one complete rewrite after they learned what they did wrong the first time...

I just gave you an example, but feel not to believe me

And I may suspect you know what it means to write an "incremental upate" engine, right? You can't use anything any longer, not even the move generator. So you start from scratch.

And BTW, for clearness sake, you started this discussion with the word "ENGINE". That excludes own interfaces, protocols like WB or UCI.

Why would that be? Cray Blitz was a heavily incremental update program. It updated pawn structure counters (isolated, doubled, passed pawns) and such, move generation information and such. I removed the incremental stuff in the 80's for simplicity and I didn't rewrite everything. Perhaps it depends on how the code is designed and written, but for me it wasn't an issue.

Engine includes whatever plays chess. From book moves, to hashing, to evaluation, to searching, to making/unmaking moves, to detecting check, generating moves, you name it. My basic move generator hasn't changed in a LONG time, even though the methodology for generating moves has. And for clarity, I did talk about "COMPLETE rewrite from scratch" which would be necessary to lose parts of fruit in later rybka versions. Others jumped in with the bogus stuff about glaurung and such, based on half-assed interpretations of what an author wrote / meant and then bending that interpretation to try to make a point that can't be made realistically.

I will elaborate a bit since we are having a misunderstanding what I mean with "incremental update" of EVAL. In a normal chess engine (depending on the size of EVAL the engine is in EVAL for 50-90% of the cpu time.

In an "incremental update" engine you don't evaluate positions but moves. You modify the eval score with the changes a move causes. And it's super fast. I did this in my first chess engine (1980) in BASIC, it could even (and was more or less obliged since the absence of a QS) evaluate if a checking move was a checkmate.

I guess you will now understand the implications of such an approach.

Who the hell cares about any of this? If Rajlich came back on to the scene tomorrow and presented a Rybka UCI updated version, that he stated was competitive with the now top three chess engines-I'd be online with Pay Pal.

And none of Dr. Hyatt's repetitious guilty tripping about Vas being a code copying cheat would make the slightest difference in my decision to do otherwise.

Enough is enough!

[Terry McCracken]

I wasn't talking about Rybka, just highlighting your condemning and patronizing remarks, the red above.

Again, eh? I have seen ZERO cases of a complete rewrite. Throw out the old code and start from scratch with nothing. That is a dumb way of developing software and that is not a "condescending remark" it is a simple statement of fact. If someone DID rewrite something N times from scratch, they are really very poor programmers. Once, perhaps, to change language, data structures, board representation and everything else. But more than once? What a waste...

So by your logic Fabien and Marcel are dumb programmers, bravo Bob.

Many realities out there but only the Hyatt reality counts.

Here is another reality, mine. Somewhere in the 90's I wrote an experimental version from scratch. The basic ideas;

1. Incremental update of the evaluation function (gain speed).
2. Don't generate all moves at once, just one at the time when you need one (gain speed).

Made no sense to use the regular Rebel code because it was fully ASM, such a fundamental different approach one does in C first and when it has a future you convert the critical parts to ASM for the final speedup.

All the reasons in the world to start from scratch.

I have YET to see where anyone has rewritten from SCRATCH. Not once. So no "bravo, bob" but rather "bravo Ed for wanting to divert the topic rather than paying attention to what has been claimed." "from scratch" means ZERO code (none, nada, etc) retained from previous version. Doesn't happen very often. Certainly not more than once for a single programmer.

Try again, failure here.

Again, the premise has been "more than once". I'll give anyone one complete rewrite after they learned what they did wrong the first time...

I just gave you an example, but feel not to believe me

And I may suspect you know what it means to write an "incremental upate" engine, right? You can't use anything any longer, not even the move generator. So you start from scratch.

And BTW, for clearness sake, you started this discussion with the word "ENGINE". That excludes own interfaces, protocols like WB or UCI.

Why would that be? Cray Blitz was a heavily incremental update program. It updated pawn structure counters (isolated, doubled, passed pawns) and such, move generation information and such. I removed the incremental stuff in the 80's for simplicity and I didn't rewrite everything. Perhaps it depends on how the code is designed and written, but for me it wasn't an issue.

Engine includes whatever plays chess. From book moves, to hashing, to evaluation, to searching, to making/unmaking moves, to detecting check, generating moves, you name it. My basic move generator hasn't changed in a LONG time, even though the methodology for generating moves has. And for clarity, I did talk about "COMPLETE rewrite from scratch" which would be necessary to lose parts of fruit in later rybka versions. Others jumped in with the bogus stuff about glaurung and such, based on half-assed interpretations of what an author wrote / meant and then bending that interpretation to try to make a point that can't be made realistically.

I will elaborate a bit since we are having a misunderstanding what I mean with "incremental update" of EVAL. In a normal chess engine (depending on the size of EVAL the engine is in EVAL for 50-90% of the cpu time.

In an "incremental update" engine you don't evaluate positions but moves. You modify the eval score with the changes a move causes. And it's super fast. I did this in my first chess engine (1980) in BASIC, it could even (and was more or less obliged since the absence of a QS) evaluate if a checking move was a checkmate.

I guess you will now understand the implications of such an approach.

Who the hell cares about any of this? If Rajlich came back on to the scene tomorrow and presented a Rybka UCI updated version, that he stated was competitive with the now top three chess engines-I'd be online with Pay Pal.

And none of Dr. Hyatt's repetitious guilty tripping about Vas being a code copying cheat would make the slightest difference in my decision to do otherwise.

This is what is wrong with this world.

[Darkmoon]

This is what it means to get over it. To let go, and to move on.

You've been in this room for far too long and all you've done is move the furniture around. That's it!

You've tried so hard to convince the others of your argument - only so that you inturn can be convinced that what you did to this man was justified and right. How bloody sad!

You're delusional if you think-even for a moment, that by having ripped this mans life apart , you've risen to a higher moral ground.