Komodo-TCEC

[lkaufman]

Thanks Larry. One suggestion is to have Komodo automatically use real threads rather than hyperthreads. Houdini 3 does it.
I have a quad Laptop and when I disable hyper it goes to 6 instead of 4. I have no problems with my desktops.
I mention this since I changed Komodo6 to 4 cores in parameters and after 14 games against Stockfish 32-bit 1 core laptop its dead even.
Komodo got blown out twice in less than 40 moves. Did not seem right. When I replayed but changed parameter to 8 cores including hyper they are draws. It seems stronger with 8. Depth is 1 deeper and total nodes is 75% greater.

With hyperthreading turned off you should certainly just use the number of real cores. With hyperthreading on there seems to be some controversy over whether to use 4 or 8 threads (for an i7 quad). It may even depend on the specific engine. Using 8 threads certainly utilizes the resources much better, but MP performance going from 4 to 8 threads is not wonderful, so it's a close call I think. There is some evidence that Komodo benefits a bit more than other engines going up from 4 cores to 8 or 16, so it is possible that setting threads to 8 might be best in Komodo with HT on. You can't tell by number of nps though.

[carldaman]

Hi Larry
Will this be a free update to those who already purchased K6? I am sorry if this was already mentioned anywhere...
Let me make it clear that I have no intentions to demand it as such. Free for K6 buyers or not, my respect to DD will be the same.
Regards

No, it certainly won't be free as it is a major upgrade. We are discussing whether to offer a discount to K6 purchasers due to the relatively short time interval between the two releases.

I would think loyal Komodo6 buyers might like a nice discount since the new release is coming out so soon after it.

For future versions, Komodo should benefit greatly if a learning or permanent hash feature is added. A lot of people buy Komodo to use for analysis, making a learning file extremely useful. The best such impementation that I know is done by Jeremy Bernstein for his version of Stockfish. It would add a lot of value to Komodo and should also justify a higher price.

Regards,
CL

[lkaufman]

Hi Larry
Will this be a free update to those who already purchased K6? I am sorry if this was already mentioned anywhere...
Let me make it clear that I have no intentions to demand it as such. Free for K6 buyers or not, my respect to DD will be the same.
Regards

No, it certainly won't be free as it is a major upgrade. We are discussing whether to offer a discount to K6 purchasers due to the relatively short time interval between the two releases.

I would think loyal Komodo6 buyers might like a nice discount since the new release is coming out so soon after it.

For future versions, Komodo should benefit greatly if a learning or permanent hash feature is added. A lot of people buy Komodo to use for analysis, making a learning file extremely useful. The best such impementation that I know is done by Jeremy Bernstein for his version of Stockfish. It would add a lot of value to Komodo and should also justify a higher price.

Regards,
CL

Yes, a nice discount for K6 buyers is planned. We can talk to Jeremy about permanent hash, he is very helpful to us.

[OneTrickPony]

Yes, a nice discount for K6 buyers is planned. We can talk to Jeremy about permanent hash, he is very helpful to us.

I am happy to see the release!
May I suggest that you guys need some consistency/clarity about release policy though ? For example now I would be happy to buy Komodo TCEC but then I am thinking maybe K7 is just around the corner and it's better to wait for that. If K7 comes soon people who didn't wait are going to feel bitter, if it doesn't come they are going to regret not buying but then will be like "I waited so long let's wait more". This is never ending cycle of guessing which causes unhappiness of buyers and I think makes some of them to not buy or buy less frequently.
I think this is very serious marketing problem of Komodo team, Don even hinted at receiving some hate mail on new release...

Something like "release now, next one planned in 6 months" or maybe release now, buying gives you upgrades in x months or even consistent policy ala Houdini: release once a year/6months and offer/don't offer discounts for middle year release would go long way in creating some trust/happiness among your clients.

[lkaufman]

Yes, a nice discount for K6 buyers is planned. We can talk to Jeremy about permanent hash, he is very helpful to us.

I am happy to see the release!
May I suggest that you guys need some consistency/clarity about release policy though ? For example now I would be happy to buy Komodo TCEC but then I am thinking maybe K7 is just around the corner and it's better to wait for that. If K7 comes soon people who didn't wait are going to feel bitter, if it doesn't come they are going to regret not buying but then will be like "I waited so long let's wait more". This is never ending cycle of guessing which causes unhappiness of buyers and I think makes some of them to not buy or buy less frequently.
I think this is very serious marketing problem of Komodo team, Don even hinted at receiving some hate mail on new release...

Something like "release now, next one planned in 6 months" or maybe release now, buying gives you upgrades in x months or even consistent policy ala Houdini: release once a year/6months and offer/don't offer discounts for middle year release would go long way in creating some trust/happiness among your clients.

Basically our policy is to release whenever we have gained about 25 elo from the previous version at 2' + 1". It is very hard to predict how long that will take. It could take many months, or we could make some huge discovery that gains that much overnight. I like to think of it as something like charging $2 per elo point gained. If the time interval is short we'll offer appropriate discounts. Hypothetically, suppose the day after we release TCEC we make a breakthru and gain another 25 points. We can then offer a new version just a week later, but charge more for it to cover the 25 elo points and offer enough discount for the recent purchasers that they won't end up paying too much more than if they had waited. Trying to schedule future releases ignores the great uncertainty about the future, especially now that we have a new chief programmer.

[Dr.Wael Deeb]

Yes, a nice discount for K6 buyers is planned. We can talk to Jeremy about permanent hash, he is very helpful to us.

I am happy to see the release!
May I suggest that you guys need some consistency/clarity about release policy though ? For example now I would be happy to buy Komodo TCEC but then I am thinking maybe K7 is just around the corner and it's better to wait for that. If K7 comes soon people who didn't wait are going to feel bitter, if it doesn't come they are going to regret not buying but then will be like "I waited so long let's wait more". This is never ending cycle of guessing which causes unhappiness of buyers and I think makes some of them to not buy or buy less frequently.
I think this is very serious marketing problem of Komodo team, Don even hinted at receiving some hate mail on new release...

Something like "release now, next one planned in 6 months" or maybe release now, buying gives you upgrades in x months or even consistent policy ala Houdini: release once a year/6months and offer/don't offer discounts for middle year release would go long way in creating some trust/happiness among your clients.

Basically our policy is to release whenever we have gained about 25 elo from the previous version at 2' + 1". It is very hard to predict how long that will take. It could take many months, or we could make some huge discovery that gains that much overnight. I like to think of it as something like charging $2 per elo point gained. If the time interval is short we'll offer appropriate discounts. Hypothetically, suppose the day after we release TCEC we make a breakthru and gain another 25 points. We can then offer a new version just a week later, but charge more for it to cover the 25 elo points and offer enough discount for the recent purchasers that they won't end up paying too much more than if they had waited. Trying to schedule future releases ignores the great uncertainty about the future, especially now that we have a new chief programmer.

That's why I suggested to gain a nice 40-45 Elo over Komodo 6 and then release it as a stand-alone Komodo 7....

You're not far from this target,maybe a couple of months and bingo regards,
Dr.D

[OneTrickPony]

We can then offer a new version just a week later, but charge more for it to cover the 25 elo points and offer enough discount for the recent purchasers that they won't end up paying too much more than if they had waited. Trying to schedule future releases ignores the great uncertainty about the future, especially now that we have a new chief programmer.

I understand that point I even see it's rational/honest policy which should be for the good of the users (they get the goodies once they are available).
There is however human nature which isn't rational. I am sure you have noticed not only some backslash but also a lot of requests/demands/entitlements for freebies. Even if people don't ask for those they will always be bitter if they bought just beyond the freebie/discount date.

I tried to put a summary of feedback I saw about Komodo. I think you are making a marketing mistake not scheduling releases and/or having clear rules who/when qualify for freebie/discount. I think you under appreciate irrational nature of a buyer and lose sails and trust with that.
I appreciate your answer and I am not going to argue the point further I hope you find my point of view useful as it represents sentiment of significant group of your potential and actual clients.

If you stick to what you are doing now I think one big improvements would be clear policy of discounts and not doing it at the time of next release. If something is clear and known from the get go people are less likely to complain about it.

[Lion]

it make sense to me.

thanks

[lkaufman]

We can then offer a new version just a week later, but charge more for it to cover the 25 elo points and offer enough discount for the recent purchasers that they won't end up paying too much more than if they had waited. Trying to schedule future releases ignores the great uncertainty about the future, especially now that we have a new chief programmer.

I understand that point I even see it's rational/honest policy which should be for the good of the users (they get the goodies once they are available).
There is however human nature which isn't rational. I am sure you have noticed not only some backslash but also a lot of requests/demands/entitlements for freebies. Even if people don't ask for those they will always be bitter if they bought just beyond the freebie/discount date.

I tried to put a summary of feedback I saw about Komodo. I think you are making a marketing mistake not scheduling releases and/or having clear rules who/when qualify for freebie/discount. I think you under appreciate irrational nature of a buyer and lose sails and trust with that.
I appreciate your answer and I am not going to argue the point further I hope you find my point of view useful as it represents sentiment of significant group of your potential and actual clients.

If you stick to what you are doing now I think one big improvements would be clear policy of discounts and not doing it at the time of next release. If something is clear and known from the get go people are less likely to complain about it.

Our policy is to offer the purchasers of the latest version a discount on the new one, the amount of which depends on the time interval (short interval means big discount). That way you can be fairly sure that if you buy a new version when it comes out, whenever the new one comes out if you upgrade you will not have paid an unreasonable amount per month for the previous one.

[Laszlo Gaspar]

What about simply Komodo Champion? (referring indirectly to Don who finally reached the top and became champion with his creature)