Joined
Last Online
Recent Posts
posted in BB2 - Ideas & Suggestions read more

This may sound a bit messy, but you can create up to 5 accounts for each bb2-key you have. So the two of you could create 10 accounts -> create a team in each and create a league with them -> share accounts and passwords for the teams depending on who has to play

posted in BB2 - Ideas & Suggestions read more

Remark: I was able to reproduce it now, thanks for your clarification.

posted in BB2 - Ideas & Suggestions read more

@voodoomike said in For the Nth time - LOWER TV Diff - Even optionaly:

@arne said in For the Nth time - LOWER TV Diff - Even optionaly:

I then filtered out draw games and games, with an absolute tvdifference smaller than 300

There's your problem. There's no reason to remove draw games in this case - we only need to do that when we're checking how many games we accurately predict the winners of, since draw games have no winner and thus don't really tell us anything about how well a metric predicted that match. The metrics don't try to predict draws.

Fair enough, how did you include draws? did you do it like this:
loss: 0
draw: 0.5
win: 1
?

I still can't reproduce your results if I include the draw games like this.

@voodoomike said in For the Nth time - LOWER TV Diff - Even optionaly:

I don't know what you mean by that. I'm generally NOT using only one datapoint per match unless we're talking about match effects that are isolated from their relationship to team specifics - there are two teams in each match, and each team has its own outcomes and data. Not considering both outcomes is ignoring half the data.

We have some serious communication issues, I could have sworn, that you critized me last time for using 2 data points per match. Anyway, I am happy, that it is an non-issue then.

edit: did you use the pearson correlation?

edit2: silly me, I overlooked something, forget what I wrote first.

posted in BB2 - Ideas & Suggestions read more

Edit: The issue was resolved and it is working now (see following posts)

Just as a small follow up to our last discussion (mean tv-difference different to 0), I tried to reproduce your numbers, but I can't. Could you check, where the differences are or what I am missing?

For the sake of similarity, I will just look at this value now:

@voodoomike said in For the Nth time - LOWER TV Diff - Even optionaly:

Beyond the 300 TV difference

Correlation between TV advantage and match outcome: r = 0.139, p < 0.01

The first steps I did:

  1. I downloaded the data base for CCL-S3 from goblinspy
  2. I exported the table teammatches as csv and read it into R (variable teammatches)

The dimensions of the variable are here:

dim(teammatches)
[1] 37834 85

I then filtered out draw games and games, with an absolute tvdifference smaller than 300

teammatches = teammatches[

teammatches[,"draw"] == 0 &
teammatches[,"tvdiffabs"] >= 300

,]
dim(teammatches)
[1] 4492 85

The correlation (pearson) between tv-difference and outcome (win) is then:

cor(teammatches[,"tvdiff"], teammatches[,"win"])
[1] 0.3545435

In this case every game is counted twice (both directions). To only include every game only once, I looked at every second row (2,4,6,...) in the table teammatches:

index = (1:(nrow(teammatches)/2))*2
sum(teammatches[index,"id"] != teammatches[index-1,"id"]) #(just a double check for this apporach)
[1] 0
cor(teammatches[index,"tvdiff"], teammatches[index,"win"])
[1] 0.2648016
mean(teammatches[index,"tvdiff"])
[1] -226.5093

I wanted to try out other things on how to select the directions. The next attempt was to only look at the perspective from the team with the higher TV

filter = index - 1*(teammatches[index,"tvdiff"] < 0)
cor(teammatches[filter,"tvdiff"], teammatches[filter,"win"])
[1] 0.07901165
mean(teammatches[filter,"tvdiff"])
[1] 389.3945

In the last attempt, I choose randomly which direction to pick

filter = index - 1*(runif(nrow(teammatches)/2)>0.5)
cor(teammatches[filter,"tvdiff"], teammatches[filter,"win"])
[1] 0.3544961
mean(teammatches[filter,"tvdiff"])
[1] 2.831701

-> The resulting correlation heavily depends on direction which was choosen. Based on our last discussion, I know that you only look at one direction and that you have somehow a systematic of choosing that direction (your mean tv-difference is different from zero). So my question would be: How do you choose the direction you look at? Otherwise: Do I have some mistakes/errors or something else in my script?

posted in BB2 - Ideas & Suggestions read more

Is there a reason for that?

I played some games in CCL recently and it seems, that there are quite a lot of noobs in CCL. This isn't a problem, if the noobs know the difference between CCL and COL and willingly decide to play CCL. However, this doesn't seem to be the case for all: I just had a game against a coach with the sad game history of 2-1-12. He seemed to be rather frustrated with the skill-level of his opponents. Thus, I had to explain him the difference between CCL and COL and warn him, that his team will get banned soon (since he had more than 5 concessions).

Overall, I think it would be better if COL was the first competition in the list to avoid such frustrating moments (games & bans) for newer players.

posted in BB2 - Ideas & Suggestions read more

@dode74 said in Change to the way teams are ranked and matched in Champions Ladder.:

And yet the current system is proving better. I've tried adjusting rankpoint gain by win probability based on TV difference, but that's still worse (although only slightly) than the current ranking system.

To add to for @Woofbark :
There should be an ELO-style system, which performs better than the current system, the problem is, however, that it is quite difficult to find that system. If you are good with data analysis / statistics, you can try to find it by yourself (do you know that you can dowanload the data from goblinspy)?

posted in BB2 - General Discussion read more

@mordrek said in goblinSpy - league stats without upload needed:

Got a request for matchup-pool-info.
It's now available as new queries under Queries->League results (click the '>>' button inside the League results button).

"Pools" - Shows the matchups with a new column "pool" that is all matchups made within a minute (so the same pool) together with TV for the teams and absolute TV difference.

"Per pool" - Shows number of matchups for each poolcount.
Poolcount is the number of teams matched in a pool, so 2 if only one match was made in that pool.

Note: You need to remove cached scripts to get this (for chrome , do a shift-click on the reload button).

If you got requests for improvements for this or other features, let me know and perhaps I'll bother.

This is pretty awesome, thank you for all your work 🙂

posted in BB2 - General Discussion read more

@mori-mori said in Matchmaking new season:

Not sure what to mind about occasional 400TV difference.[...].

This depends on your motivation for playing the game. If you just play it for fun and don't care so much about winning, then yes, the matches with a high TV-differences wouldn't be a problem. However, coaches who are also interested in their overall ranking (which is not uncommon in CCL), can rightfully get frustrated, when they have to fight uphill battles and yes matches in which they are in a TV-disadvantage are often such uphill battles. Same goes with coaches, who want to develop their team, since the higher developed teams often have more tools to damage their team.

Personally, I am not really happy with the ranking/matchmaking in CCL either, it is just so dependent on the opponents which you get (especially with the introduction of the lower-tier-teams). The best example of what I mean would probably be the amazon-coach, who prays every time that he won't get matched against a dwarf. Unfortunately, I don't see really the perfect solution for it: In terms of matchmaking, the player base seems to be too small to guarantee a good match-up all the time and in terms of ranking: It would need a perfect elo-like system to tackle these issues, but as far as I know, this doesn't exist and I doubt, that it can't be easily created. So I guess the system we have, is almost the best we can get with the tools available

But hey: I still like the game and I will still continue playing it, I will just wait for the next season before I go back to it =).

@dode74 said in Matchmaking new season:

[...] (presumably due to the spam filters - I've had the same issue).

Just write an additional sentence at the end of a post to make it acceptable for the spam-filter

posted in BB2 - Ideas & Suggestions read more

@ugh said in Max 12 or 13 losses in the Championship qualifier:

Is the API missing to do this? So many things can be queried by now, why is there no way to call a function that disqualifies a team/bans a player from a competition? Is that a suggestion that fell through? If so, why? If it can be done manually, providing an API which has the same functionality as the reaction to someone using some buttons in the GUI (i.e. a functionality that is implemented already) should not be a monumental task.

Didn't we talk about this before? Currently, everything (including the banning of coaches who conceded more than 5 times) is done manually via a web interface in a very inefficient manner. It may (or may not) be possible to optimize the administrative work by quite a lot if some hero-programmer writes additional tools for the admins to make the steps more automatic. If you feel like you can be that hero-programmer, grab a cape and talk to Dode, he may give you more information about it. (I am serious about that, I even asked Dode the last time, whether it is ok, if I make such advertisement for the admin-team).


Looks like your connection to Focus Home Interactive - Official Forums was lost, please wait while we try to reconnect.