|
On July 17 2013 09:59 Entirety wrote:
Can you make a feature where Aligulac actually picks a result?
Say, for example, that there is a match. Player A vs. Player B
Player A 2-0 30%
Player A 2-1 40%
Player B 2-1 20%
Player B 2-0 10%
Then Aligulac actually picks one of those results, so that there is a 30% chance that it chooses Player A 2-0, 40% chance that it chooses Player A 2-1, etc. Then, if you try again, then Aligulac might pick a different result.
I wrote a script for this using Python and BeautifulSoup, it needs a bit of cleaning up but it wasn't very difficult to get the results. Will probably post it in the next few days.
|
|
|
First of all, thank you for all the hard work you've done on this amazing website. It's the best predication tool I've seen.
I was browsing through the website when I noticed some of the adjustments didn't seem to make sense. I noticed this explanation in the FAQ:
"The upshot of this is that if a player overperforms versus Terran (say 2–0 when 1–1 was expected), but significantly underperforms in the other matchups (say 0–10 in each when 5–5 was expected), the rating versus Terran may still decrease."
Will this still happen for low uncertainty?
If we have a hypothetical player with 100% winrate versus Terran but has 0% winrate in the other two (and perhaps is matched up against Terran less) will their vT rating just continually decrease?
Edit: Will there a periodic report on the website on accuracy of predictions? 
|
First off great website with great simple design and stacked full of stats :D
I have a suggestion for another column ranking in the teams section. The Proleague and All-kill scores are a good historical reference but I am always wanting to know who the best teams are right now.
So I suggest a third column where the teams can be ranked by the average rating points of their top 5 players (5 players with currently highest rating points). I see there is already room for another column, could be called 'Top 5 Players' or something.
|
On August 26 2013 14:21 ThunderGod wrote:
First off great website with great simple design and stacked full of stats :D
I have a suggestion for another column ranking in the teams section. The Proleague and All-kill scores are a good historical reference but I am always wanting to know who the best teams are right now.
So I suggest a third column where the teams can be ranked by the average rating points of their top 5 players (5 players with currently highest rating points). I see there is already room for another column, could be called 'Top 5 Players' or something.
Scores are not historical measurement but projection (albeit very bad ones) of a team ability to win in these formats.
Nonetheless I added your request : https://github.com/TheBB/aligulac/issues/85
|
About the match entries,since sometimes the aligulac staff forget some (well its normal they're human after all) i sometime enter some.
I would like to know if there is an easier way to make the ro32/16/8/4/2 thing.
I mean look at this :
http://i.imgur.com/PQPtF5P.png?1
Now to enter the Ro16 results i have to send a whole new request again ?
I mean its ok but it takes more time,my suggestion would be adding some kind of system that allows us to type multiple rounds at once.
|
Feel free to type in all the rounds in a single request, so long as you write in the notes explicitly which matches belong to which rounds. (First 16, then next 8, then next 4, etc.) That's the best I can do with the current system, I'm afraid.
|
Oh so i can do everything in a single request ?
That's gonna save some time thanks.
I can just put a blank line between different rounds and explain this in notes ?
|
Explaining this in notes should be fine, as the matches/events can always be changed later. If you link to Liquipedia, it's usually obvious what round a match was played in anyhow.
|
On September 04 2013 04:19 shid0x wrote:
I can just put a blank line between different rounds and explain this in notes ?
Blank lines are just skipped and can't be seen by us on the other end, you actually have to count. :/
Edit: Or as Conti says, you can just let us figure it out.
|
KeSPA released their ten-year log of player records.
If we were prone to wasting a good amount of time, it would be fun to walk these logs to recreate a timeline for players' skills in Broodwar.
|
On September 07 2013 04:15 WigglingSquid wrote:
KeSPA released their ten-year log of player records.
If we were prone to wasting a good amount of time, it would be fun to walk these logs to recreate a timeline for players' skills in Broodwar.
Would be amazing :o
|
On September 07 2013 04:15 WigglingSquid wrote:
KeSPA released their ten-year log of player records.
Hmm, link? 
|
|
|
Tlpd contains then already I think.
|
On August 26 2013 16:09 Otolia wrote:Show nested quote +On August 26 2013 14:21 ThunderGod wrote:
First off great website with great simple design and stacked full of stats :D
I have a suggestion for another column ranking in the teams section. The Proleague and All-kill scores are a good historical reference but I am always wanting to know who the best teams are right now.
So I suggest a third column where the teams can be ranked by the average rating points of their top 5 players (5 players with currently highest rating points). I see there is already room for another column, could be called 'Top 5 Players' or something. Scores are not historical measurement but projection (albeit very bad ones) of a team ability to win in these formats. Nonetheless I added your request : https://github.com/TheBB/aligulac/issues/85
Thanks, to clarify I meant average rating of top 5 players, not the whole team.
I would like to see this for countries too so we can see the relative strength of different countries with a number attached to it. We could see the depth of a country and it's likelihood to win a match against any other country, same as for team battles. For example average rating of top 5 players in
Korea: 1926
Canada: 1460
Sweden: 1452
Ukraine: 1388
USA: 1321
These kind of stats are interesting to me
|
On September 07 2013 04:15 WigglingSquid wrote:
KeSPA released their ten-year log of player records.
If we were prone to wasting a good amount of time, it would be fun to walk these logs to recreate a timeline for players' skills in Broodwar.
I believe it were only proleague matches released.
|
Hey,
Does the prediction tool use the "preview" ratings (i.e does it take into account games that were played a day or two before the match by each of the players), or does it use the previous list's rating?
|
On September 15 2013 21:20 CtrlAltDefeat wrote:
Hey,
Does the prediction tool use the "preview" ratings (i.e does it take into account games that were played a day or two before the match by each of the players), or does it use the previous list's rating?
Pretty almost sure that it uses the preview matches.
Edit: It uses the preview rating. Currently that means all matches except DH Ro16 (excluding Stardust-Elfi). The preview is updated every 6 hours. Currently that means it will be updated in 2½ hours.
|
Big update today. I'm tired of writing changelogs, so I'll just copy the brief one:
- Player, event and rating list pages now have some more info shown on top in a tabbed box.
- Player rating charts now show the number of games played per period.
- Predictions (now under the inference menu) for single matches (shows more data) and proleague matches (you can now simulate ace matches as they should be) have been improved.
- The team transfer page looks better than ever before.
- You can now navigate directly to any submenu item by hovering over the relevant entry. (This is a bit tricky though… they're a bit small.)
- The team list now shows the average rating of the top five players, as well as the number of players.
- Many small things…
Mostly, the exciting improvements are under the hood:
- Now using Python 3.3, Django 1.6 and PostgreSQL.
- The event hierarchy table has been very much improved and should now be much more stable and fast. However, this has introduced a bit of difficulty with regards to sorting. For this reason you might notice that matches are sorted in a funny order sometimes. I will fix this, don't worry.
- Tons of bugfixes.
I have decided to remove the compare feature, since it was pretty dull and mostly superseded by the match prediction page. I can add the p-value there instead, if anyone's interested.
In the beginning, there will likely be some problems and bugs owing to the rewrite. Please feel free to report issues here.
For those of you who want SSL support, yeah I've been looking at it and I think I can manage.
|
|
|
|
|
|
EPT
