new calc method in Tactics Trainer

Sort:
carpon

Has anyone noticed a new method of calculation of ratings in the Tactics Trainer? Less points allowed for success, less taken off for failure. I won't be ranging 200 points from one day to the next but it seems much harder to get a run going.

DeepGreene

No, I hadn't noticed...  How did you figure this out?

It almost sounds like a kind of Tactics Trainer RD value manifesting.  But I'm not sure TT ratings use any RD (?).  I tend to think that they shouldn't, in any case.

erik

i think we did change the RD to make it less volatile

EternalChess
erik wrote:

i think we did change the RD to make it less volatile


 You think?!

 Its nice to know that the owner of Chess.com knows what is going on in chess.com.

DeepGreene
SerbianChessStar wrote:
erik wrote:

i think we did change the RD to make it less volatile


 You think?!

 Its nice to know that the owner of Chess.com knows what is going on in chess.com.


Some delegation of responsibility is a necessity for any successful leader.  :)

slack

I like the new change. Instead of being penalized -40 or so points for using too much time on a problem I passed, I only lose about 15.

This sends the right message. Players should be encouraged to study the board and see the variations before making the move.

ASpieboy

You shouldn't lose points on a problem you passed PERIOD!

carpon

Generally I approve of the change for the same reasons listed above. It will have some implications for the future. For example, a person's rating at any given time wil be more indicative of where they "should be" rated. Also, it will take more problems to get to the top. Lets pay Magnus Karlson $100 to see how long it takes him to get to 2000. It might also fix the incongruity of some problems I noted in my blog - the ones that seem rated too low.

EternalChess
carpon wrote:

Generally I approve of the change for the same reasons listed above. It will have some implications for the future. For example, a person's rating at any given time wil be more indicative of where they "should be" rated. Also, it will take more problems to get to the top. Lets pay Magnus Karlson $100 to see how long it takes him to get to 2000. It might also fix the incongruity of some problems I noted in my blog - the ones that seem rated too low.


 it takes me 3 minutes to get to 2000..

Lets see if Magnus Carlsen can hit 3000+!

kunduk

hmm..

Mihauki

Hi, I think the players list for TT rating doesn't work today. You cannot find any players and you cannot sort the players by ratings or any other order.

Yesterday and all previous days it worked fine. I have a bet going on with my brother who has the higher rating by the end of each month so I like to check out where he's at :-)

I also noticed the new calc method. The player with 4000 points got deleted, but overall the ratings have grown. It use to be enough to have 2570 or so to be in the top 100 now you need 2630 or so. The ratings should be more realistic. On another TT site I have a rating of 1960 here I got 2500 and I'm sure the first one is quite accurate.

Besides I think now you get less points taken away if you have some of the first moves right. Lets say if you got 5 out of 7 right you lose 10 points, even thought you lose the game. But if you make another first move that only gets a pawn less compared to the right solution you get minus 40 points. I think it sould be the other way.

carpon
SerbianChessStar wrote:
carpon wrote:

Generally I approve of the change for the same reasons listed above. It will have some implications for the future. For example, a person's rating at any given time wil be more indicative of where they "should be" rated. Also, it will take more problems to get to the top. Lets pay Magnus Karlson $100 to see how long it takes him to get to 2000. It might also fix the incongruity of some problems I noted in my blog - the ones that seem rated too low.


 it takes me 3 minutes to get to 2000..

Lets see if Magnus Carlsen can hit 3000+!


I wasn't very clear. What I meant was, how long would it take a really good player to start at 1200 and advance through the problems to 2000. The best you can get, I should say, I can get, is +15. If you were blazingly fast you might get 18, but lets call it 20. The diff between 1200 and 2000 is 600 pts. If you got every one in the least possible time it would take a minimum of 30 problems to get to 2000. So how long would that take. We need a volunteer Grand Master who has not used chess.com.

erik

i think your own tactics RD should start out higher and then get smaller.

tunatin, sorry, no list right now :(

carpon
Mihauki wrote:

Hi, I think the players list for TT rating doesn't work today. You cannot find any players and you cannot sort the players by ratings or any other order.

Yesterday and all previous days it worked fine. I have a bet going on with my brother who has the higher rating by the end of each month so I like to check out where he's at :-)

I also noticed the new calc method. The player with 4000 points got deleted, but overall the ratings have grown. It use to be enough to have 2570 or so to be in the top 100 now you need 2630 or so. The ratings should be more realistic. On another TT site I have a rating of 1960 here I got 2500 and I'm sure the first one is quite accurate.

Besides I think now you get less points taken away if you have some of the first moves right. Lets say if you got 5 out of 7 right you lose 10 points, even thought you lose the game. But if you make another first move that only gets a pawn less compared to the right solution you get minus 40 points. I think it sould be the other way.


I just did one 3/5 moves right and only got 1 pt.

I think the new system is more difficult becuz the most points I've ever received is 14 but I lose 15,16 for 0%. It should be possible to get the maximum points. And I agree with the player who said you shouldn't lose pts if you get it right. Another change I noticed is that more points are removed from a problem than what you receive. Over time that will reduce the trainee ratings, which is what mihauki suggested.

flamencowizard

I wonder how the system rounds off the numbers in its calculations.  Does it round to the nearest, or round down or up?  The way it rounds numbers off could have a greater influence now that the volitility is down.