|
Dakota_Fanning
Hungary2371 Posts
🎂 repmastered.app celebrates its first birthday 🎉, let's dive into the numbers of Millions.
One Million Replays uploaded Half Million 👨Players in the Database Two Million 👬Player Connections (Unique Player pairs) Nearly 0.01 Million 🗺️Maps (Ten Thousand) One Million ▤Map Leaderboard Entries (Unique Map-Player pairs) Starting from 2001 there are 135 new replays every single day (on average). Or in other words: there's a new repay in every 10 minutes (starting from 2001).
To see exact and up-to-date numbers and even more stats, visit the new repmastered.app global stats page.
You can find this link in the footer of the site.
Thanks a Million for the Million replays! 👏
Let's hope the 2nd year brings another million! I encourage everyone to upload all your replays before they are gone forever. No registration is required. Become part of History, become part of StarCraft's Legacy!
|
Congrats on your database!
To contribute, i uploaded 50k replays, out of which only about 15k were not duplicates. Still something.
"...replays do not store the information who won. I wrote some algorithms to detect winners based on what’s available in replays, but it never will be 100% accurate."
I did not find race winning stats in the global stats page. Could you please extract tvp, tvz, pvz winning stats overall, by year, and by main maps?
Thank you!
A few more development ideas:
- in the player search, there could be an option for "player name includes", becaues the clan tag is often before the name of the player, and this way the results don't show up.
- more than 20 hits could show up on one page, or the user should be able to modify the default.
- if there are many pages of hits, it shoud show how many.
- there should be an option to "select all" and "download all"
|
CONGRATULATIONS! To me as well for sending my reps!
|
is it possible to look at all replays under your name
|
TLADT24920 Posts
On April 18 2021 11:46 jinjin5000 wrote:
is it possible to look at all replays under your name
?
|
Dakota_Fanning
Hungary2371 Posts
On April 18 2021 06:12 zimp wrote:
Congrats on your database!
To contribute, i uploaded 50k replays, out of which only about 15k were not duplicates. Still something.
"...replays do not store the information who won. I wrote some algorithms to detect winners based on what’s available in replays, but it never will be 100% accurate."
I did not find race winning stats in the global stats page. Could you please extract tvp, tvz, pvz winning stats overall, by year, and by main maps?
Thank you!
A few more development ideas:
- in the player search, there could be an option for "player name includes", becaues the clan tag is often before the name of the player, and this way the results don't show up.
- more than 20 hits could show up on one page, or the user should be able to modify the default.
- if there are many pages of hits, it shoud show how many.
- there should be an option to "select all" and "download all"
Very nice, thank you.
Overall race / matchup stats do not yet show winning stats, I will add those. All maps already include matchup stats. Look at the map profile, e.g. here's Fighting Spirit, you can see overall race stats and also broken down for each matchup played on the map.
About player search: there is already a "Anywhere in players" search type, that is equivalent to "player name includes" 
Yes, I plan to customize the page size.
Showing how many results there are requires running the query twice, increasing the server load and response time by 2.
I also plan to add support for downloading multiple reps.
|
Dakota_Fanning
Hungary2371 Posts
On April 18 2021 11:46 jinjin5000 wrote:
is it possible to look at all replays under your name
If you mean replays you uploaded, yes, select the "My Uploads" tag when searching.
If you mean all reps under a specific player, yes, enter that player name to the Players search field. If you mean games of multiple players (your akas), that's not yet supported. You can list games with multiple players, but that requires all players being in the same game. You can however view aggregated statistics of all your players, just list them all separated by comma on the Players page.
|
Dakota_Fanning
Hungary2371 Posts
On April 18 2021 15:50 Dakota_Fanning wrote:Show nested quote +On April 18 2021 06:12 zimp wrote:
Congrats on your database!
To contribute, i uploaded 50k replays, out of which only about 15k were not duplicates. Still something.
"...replays do not store the information who won. I wrote some algorithms to detect winners based on what’s available in replays, but it never will be 100% accurate."
I did not find race winning stats in the global stats page. Could you please extract tvp, tvz, pvz winning stats overall, by year, and by main maps?
Thank you!
A few more development ideas:
- in the player search, there could be an option for "player name includes", becaues the clan tag is often before the name of the player, and this way the results don't show up.
- more than 20 hits could show up on one page, or the user should be able to modify the default.
- if there are many pages of hits, it shoud show how many.
- there should be an option to "select all" and "download all" Very nice, thank you. Overall race / matchup stats do not yet show winning stats, I will add those. All maps already include matchup stats. Look at the map profile, e.g. here's Fighting Spirit, you can see overall race stats and also broken down for each matchup played on the map. About player search: there is already a "Anywhere in players" search type, that is equivalent to "player name includes" Yes, I plan to customize the page size. Showing how many results there are requires running the query twice, increasing the server load and response time by 2. I also plan to add support for downloading multiple reps.
OK, global race stats have been added, by matchups. Visit the Matchups page to see win ratio of all non-mirror, 2-team balanced matchups.
From that: P > T > Z
P > Z (53%)
P > T (57%)
T > Z (54%)
(Obviously this is just calculated from the uploaded 1v1 replays.)
|
Congratulations and keep up the great work Super happy to have all the replays I managed to save over the years available online, on demand.
|
Nice bro. Very cool homepage indeed. Hopefully it will remain active for quite a while and really can become a legacy archive
|
On April 19 2021 22:19 Dakota_Fanning wrote:Show nested quote +On April 18 2021 15:50 Dakota_Fanning wrote:On April 18 2021 06:12 zimp wrote:
Congrats on your database!
To contribute, i uploaded 50k replays, out of which only about 15k were not duplicates. Still something.
"...replays do not store the information who won. I wrote some algorithms to detect winners based on what’s available in replays, but it never will be 100% accurate."
I did not find race winning stats in the global stats page. Could you please extract tvp, tvz, pvz winning stats overall, by year, and by main maps?
Thank you!
A few more development ideas:
- in the player search, there could be an option for "player name includes", becaues the clan tag is often before the name of the player, and this way the results don't show up.
- more than 20 hits could show up on one page, or the user should be able to modify the default.
- if there are many pages of hits, it shoud show how many.
- there should be an option to "select all" and "download all" Very nice, thank you. Overall race / matchup stats do not yet show winning stats, I will add those. All maps already include matchup stats. Look at the map profile, e.g. here's Fighting Spirit, you can see overall race stats and also broken down for each matchup played on the map. About player search: there is already a "Anywhere in players" search type, that is equivalent to "player name includes" Yes, I plan to customize the page size. Showing how many results there are requires running the query twice, increasing the server load and response time by 2. I also plan to add support for downloading multiple reps. OK, global race stats have been added, by matchups. Visit the Matchups page to see win ratio of all non-mirror, 2-team balanced matchups. From that: P > T > Z P > Z (53%) P > T (57%) T > Z (54%) (Obviously this is just calculated from the uploaded 1v1 replays.)
Thank you!
That is impressive! This is the largest dataset from which race stats were calculated (that i know of).
Now unfortunately that confirms my guess, namely that starcraft bw is not a balanced game, if we consider all levels. It is skewed towards P.
Would it be possible to add a year filter? Or just have stats for each year somehow. To see if the evolution of the maps / strats actually changed the balance?
What algorithms do you use to determine the winner, and do you have any idea what % might be the margin of error?
|
On April 20 2021 09:40 zimp wrote:Show nested quote +On April 19 2021 22:19 Dakota_Fanning wrote:On April 18 2021 15:50 Dakota_Fanning wrote:On April 18 2021 06:12 zimp wrote:
Congrats on your database!
To contribute, i uploaded 50k replays, out of which only about 15k were not duplicates. Still something.
"...replays do not store the information who won. I wrote some algorithms to detect winners based on what’s available in replays, but it never will be 100% accurate."
I did not find race winning stats in the global stats page. Could you please extract tvp, tvz, pvz winning stats overall, by year, and by main maps?
Thank you!
A few more development ideas:
- in the player search, there could be an option for "player name includes", becaues the clan tag is often before the name of the player, and this way the results don't show up.
- more than 20 hits could show up on one page, or the user should be able to modify the default.
- if there are many pages of hits, it shoud show how many.
- there should be an option to "select all" and "download all" Very nice, thank you. Overall race / matchup stats do not yet show winning stats, I will add those. All maps already include matchup stats. Look at the map profile, e.g. here's Fighting Spirit, you can see overall race stats and also broken down for each matchup played on the map. About player search: there is already a "Anywhere in players" search type, that is equivalent to "player name includes" Yes, I plan to customize the page size. Showing how many results there are requires running the query twice, increasing the server load and response time by 2. I also plan to add support for downloading multiple reps. OK, global race stats have been added, by matchups. Visit the Matchups page to see win ratio of all non-mirror, 2-team balanced matchups. From that: P > T > Z P > Z (53%) P > T (57%) T > Z (54%) (Obviously this is just calculated from the uploaded 1v1 replays.) Thank you! That is impressive! This is the largest dataset from which race stats were calculated (that i know of). Now unfortunately that confirms my guess, namely that starcraft bw is not a balanced game, if we consider all levels. It is skewed towards P. Would it be possible to add a year filter? Or just have stats for each year somehow. To see if the evolution of the maps / strats actually changed the balance? What algorithms do you use to determine the winner, and do you have any idea what % might be the margin of error?
A lot of games have no winner reported - for example, in my head-to-head against my Terran friend between two of our accounts, it was something like I won 9 games and he won 1 game, but out of the 11 or whatever games with no win decided according to the site, he won most of them. If everyone has the same experience as we do, then Protoss would be heavily favored lol. Perhaps something to look into - if the undecided winner games favor/disfavor certain races.
|
Dakota_Fanning
Hungary2371 Posts
On April 20 2021 09:40 zimp wrote:Show nested quote +On April 19 2021 22:19 Dakota_Fanning wrote:On April 18 2021 15:50 Dakota_Fanning wrote:On April 18 2021 06:12 zimp wrote:
Congrats on your database!
To contribute, i uploaded 50k replays, out of which only about 15k were not duplicates. Still something.
"...replays do not store the information who won. I wrote some algorithms to detect winners based on what’s available in replays, but it never will be 100% accurate."
I did not find race winning stats in the global stats page. Could you please extract tvp, tvz, pvz winning stats overall, by year, and by main maps?
Thank you!
A few more development ideas:
- in the player search, there could be an option for "player name includes", becaues the clan tag is often before the name of the player, and this way the results don't show up.
- more than 20 hits could show up on one page, or the user should be able to modify the default.
- if there are many pages of hits, it shoud show how many.
- there should be an option to "select all" and "download all" Very nice, thank you. Overall race / matchup stats do not yet show winning stats, I will add those. All maps already include matchup stats. Look at the map profile, e.g. here's Fighting Spirit, you can see overall race stats and also broken down for each matchup played on the map. About player search: there is already a "Anywhere in players" search type, that is equivalent to "player name includes" Yes, I plan to customize the page size. Showing how many results there are requires running the query twice, increasing the server load and response time by 2. I also plan to add support for downloading multiple reps. OK, global race stats have been added, by matchups. Visit the Matchups page to see win ratio of all non-mirror, 2-team balanced matchups. From that: P > T > Z P > Z (53%) P > T (57%) T > Z (54%) (Obviously this is just calculated from the uploaded 1v1 replays.) Thank you! That is impressive! This is the largest dataset from which race stats were calculated (that i know of). Now unfortunately that confirms my guess, namely that starcraft bw is not a balanced game, if we consider all levels. It is skewed towards P. Would it be possible to add a year filter? Or just have stats for each year somehow. To see if the evolution of the maps / strats actually changed the balance? What algorithms do you use to determine the winner, and do you have any idea what % might be the margin of error?
Currently it's not possible to show matchup stats by year as it is not gathered separately. Might add it in the future.
The algorithm I use is rather complex, but essence of it is tracking who leaves first, and the "largest remaining team" wins. This tracking is less accurate in replays before 2009, as those replays contain less information the algorithm uses. If anyone's interested, I open sourced the replay parsing engine including the winner detection algorithm here: https://github.com/icza/screp. I have no information about the error margin. But if the algorithm cannot decide who's the winner, that replay is added to the "unknown" part of the W-L-U record, and it is not included in the ratio calculation.
|
Dakota_Fanning
Hungary2371 Posts
On April 20 2021 10:02 Jealous wrote:Show nested quote +On April 20 2021 09:40 zimp wrote:On April 19 2021 22:19 Dakota_Fanning wrote:On April 18 2021 15:50 Dakota_Fanning wrote:On April 18 2021 06:12 zimp wrote:
Congrats on your database!
To contribute, i uploaded 50k replays, out of which only about 15k were not duplicates. Still something.
"...replays do not store the information who won. I wrote some algorithms to detect winners based on what’s available in replays, but it never will be 100% accurate."
I did not find race winning stats in the global stats page. Could you please extract tvp, tvz, pvz winning stats overall, by year, and by main maps?
Thank you!
A few more development ideas:
- in the player search, there could be an option for "player name includes", becaues the clan tag is often before the name of the player, and this way the results don't show up.
- more than 20 hits could show up on one page, or the user should be able to modify the default.
- if there are many pages of hits, it shoud show how many.
- there should be an option to "select all" and "download all" Very nice, thank you. Overall race / matchup stats do not yet show winning stats, I will add those. All maps already include matchup stats. Look at the map profile, e.g. here's Fighting Spirit, you can see overall race stats and also broken down for each matchup played on the map. About player search: there is already a "Anywhere in players" search type, that is equivalent to "player name includes" Yes, I plan to customize the page size. Showing how many results there are requires running the query twice, increasing the server load and response time by 2. I also plan to add support for downloading multiple reps. OK, global race stats have been added, by matchups. Visit the Matchups page to see win ratio of all non-mirror, 2-team balanced matchups. From that: P > T > Z P > Z (53%) P > T (57%) T > Z (54%) (Obviously this is just calculated from the uploaded 1v1 replays.) Thank you! That is impressive! This is the largest dataset from which race stats were calculated (that i know of). Now unfortunately that confirms my guess, namely that starcraft bw is not a balanced game, if we consider all levels. It is skewed towards P. Would it be possible to add a year filter? Or just have stats for each year somehow. To see if the evolution of the maps / strats actually changed the balance? What algorithms do you use to determine the winner, and do you have any idea what % might be the margin of error? A lot of games have no winner reported - for example, in my head-to-head against my Terran friend between two of our accounts, it was something like I won 9 games and he won 1 game, but out of the 11 or whatever games with no win decided according to the site, he won most of them. If everyone has the same experience as we do, then Protoss would be heavily favored lol. Perhaps something to look into - if the undecided winner games favor/disfavor certain races.
It's less likely the algorithm can detect winners in reps prior to 2009 as they contain less info. And as said above, if the winner is unknown, the game is not included in win ratio.
|
On April 20 2021 15:33 Dakota_Fanning wrote:Show nested quote +On April 20 2021 10:02 Jealous wrote:On April 20 2021 09:40 zimp wrote:On April 19 2021 22:19 Dakota_Fanning wrote:On April 18 2021 15:50 Dakota_Fanning wrote:On April 18 2021 06:12 zimp wrote:
Congrats on your database!
To contribute, i uploaded 50k replays, out of which only about 15k were not duplicates. Still something.
"...replays do not store the information who won. I wrote some algorithms to detect winners based on what’s available in replays, but it never will be 100% accurate."
I did not find race winning stats in the global stats page. Could you please extract tvp, tvz, pvz winning stats overall, by year, and by main maps?
Thank you!
A few more development ideas:
- in the player search, there could be an option for "player name includes", becaues the clan tag is often before the name of the player, and this way the results don't show up.
- more than 20 hits could show up on one page, or the user should be able to modify the default.
- if there are many pages of hits, it shoud show how many.
- there should be an option to "select all" and "download all" Very nice, thank you. Overall race / matchup stats do not yet show winning stats, I will add those. All maps already include matchup stats. Look at the map profile, e.g. here's Fighting Spirit, you can see overall race stats and also broken down for each matchup played on the map. About player search: there is already a "Anywhere in players" search type, that is equivalent to "player name includes" Yes, I plan to customize the page size. Showing how many results there are requires running the query twice, increasing the server load and response time by 2. I also plan to add support for downloading multiple reps. OK, global race stats have been added, by matchups. Visit the Matchups page to see win ratio of all non-mirror, 2-team balanced matchups. From that: P > T > Z P > Z (53%) P > T (57%) T > Z (54%) (Obviously this is just calculated from the uploaded 1v1 replays.) Thank you! That is impressive! This is the largest dataset from which race stats were calculated (that i know of). Now unfortunately that confirms my guess, namely that starcraft bw is not a balanced game, if we consider all levels. It is skewed towards P. Would it be possible to add a year filter? Or just have stats for each year somehow. To see if the evolution of the maps / strats actually changed the balance? What algorithms do you use to determine the winner, and do you have any idea what % might be the margin of error? A lot of games have no winner reported - for example, in my head-to-head against my Terran friend between two of our accounts, it was something like I won 9 games and he won 1 game, but out of the 11 or whatever games with no win decided according to the site, he won most of them. If everyone has the same experience as we do, then Protoss would be heavily favored lol. Perhaps something to look into - if the undecided winner games favor/disfavor certain races. It's less likely the algorithm can detect winners in reps prior to 2009 as they contain less info. And as said above, if the winner is unknown, the game is not included in win ratio.
Yes, I meant that certain match-ups are harder to calculate the winner for when it is one race over another, for example Protoss is easier to determine to have beaten Terran than vice versa, resulting in more PvT wins being reported and more TvP games being undecided, thus influencing MU win%.
|
Dakota_Fanning
Hungary2371 Posts
On April 20 2021 21:14 Jealous wrote:Show nested quote +On April 20 2021 15:33 Dakota_Fanning wrote:On April 20 2021 10:02 Jealous wrote:On April 20 2021 09:40 zimp wrote:On April 19 2021 22:19 Dakota_Fanning wrote:On April 18 2021 15:50 Dakota_Fanning wrote:On April 18 2021 06:12 zimp wrote:
Congrats on your database!
To contribute, i uploaded 50k replays, out of which only about 15k were not duplicates. Still something.
"...replays do not store the information who won. I wrote some algorithms to detect winners based on what’s available in replays, but it never will be 100% accurate."
I did not find race winning stats in the global stats page. Could you please extract tvp, tvz, pvz winning stats overall, by year, and by main maps?
Thank you!
A few more development ideas:
- in the player search, there could be an option for "player name includes", becaues the clan tag is often before the name of the player, and this way the results don't show up.
- more than 20 hits could show up on one page, or the user should be able to modify the default.
- if there are many pages of hits, it shoud show how many.
- there should be an option to "select all" and "download all" Very nice, thank you. Overall race / matchup stats do not yet show winning stats, I will add those. All maps already include matchup stats. Look at the map profile, e.g. here's Fighting Spirit, you can see overall race stats and also broken down for each matchup played on the map. About player search: there is already a "Anywhere in players" search type, that is equivalent to "player name includes" Yes, I plan to customize the page size. Showing how many results there are requires running the query twice, increasing the server load and response time by 2. I also plan to add support for downloading multiple reps. OK, global race stats have been added, by matchups. Visit the Matchups page to see win ratio of all non-mirror, 2-team balanced matchups. From that: P > T > Z P > Z (53%) P > T (57%) T > Z (54%) (Obviously this is just calculated from the uploaded 1v1 replays.) Thank you! That is impressive! This is the largest dataset from which race stats were calculated (that i know of). Now unfortunately that confirms my guess, namely that starcraft bw is not a balanced game, if we consider all levels. It is skewed towards P. Would it be possible to add a year filter? Or just have stats for each year somehow. To see if the evolution of the maps / strats actually changed the balance? What algorithms do you use to determine the winner, and do you have any idea what % might be the margin of error? A lot of games have no winner reported - for example, in my head-to-head against my Terran friend between two of our accounts, it was something like I won 9 games and he won 1 game, but out of the 11 or whatever games with no win decided according to the site, he won most of them. If everyone has the same experience as we do, then Protoss would be heavily favored lol. Perhaps something to look into - if the undecided winner games favor/disfavor certain races. It's less likely the algorithm can detect winners in reps prior to 2009 as they contain less info. And as said above, if the winner is unknown, the game is not included in win ratio. Yes, I meant that certain match-ups are harder to calculate the winner for when it is one race over another, for example Protoss is easier to determine to have beaten Terran than vice versa, resulting in more PvT wins being reported and more TvP games being undecided, thus influencing MU win%.
The winner detection algorithm does not depend on the races.
|
On April 20 2021 21:28 Dakota_Fanning wrote:Show nested quote +On April 20 2021 21:14 Jealous wrote:On April 20 2021 15:33 Dakota_Fanning wrote:On April 20 2021 10:02 Jealous wrote:On April 20 2021 09:40 zimp wrote:On April 19 2021 22:19 Dakota_Fanning wrote:On April 18 2021 15:50 Dakota_Fanning wrote:On April 18 2021 06:12 zimp wrote:
Congrats on your database!
To contribute, i uploaded 50k replays, out of which only about 15k were not duplicates. Still something.
"...replays do not store the information who won. I wrote some algorithms to detect winners based on what’s available in replays, but it never will be 100% accurate."
I did not find race winning stats in the global stats page. Could you please extract tvp, tvz, pvz winning stats overall, by year, and by main maps?
Thank you!
A few more development ideas:
- in the player search, there could be an option for "player name includes", becaues the clan tag is often before the name of the player, and this way the results don't show up.
- more than 20 hits could show up on one page, or the user should be able to modify the default.
- if there are many pages of hits, it shoud show how many.
- there should be an option to "select all" and "download all" Very nice, thank you. Overall race / matchup stats do not yet show winning stats, I will add those. All maps already include matchup stats. Look at the map profile, e.g. here's Fighting Spirit, you can see overall race stats and also broken down for each matchup played on the map. About player search: there is already a "Anywhere in players" search type, that is equivalent to "player name includes" Yes, I plan to customize the page size. Showing how many results there are requires running the query twice, increasing the server load and response time by 2. I also plan to add support for downloading multiple reps. OK, global race stats have been added, by matchups. Visit the Matchups page to see win ratio of all non-mirror, 2-team balanced matchups. From that: P > T > Z P > Z (53%) P > T (57%) T > Z (54%) (Obviously this is just calculated from the uploaded 1v1 replays.) Thank you! That is impressive! This is the largest dataset from which race stats were calculated (that i know of). Now unfortunately that confirms my guess, namely that starcraft bw is not a balanced game, if we consider all levels. It is skewed towards P. Would it be possible to add a year filter? Or just have stats for each year somehow. To see if the evolution of the maps / strats actually changed the balance? What algorithms do you use to determine the winner, and do you have any idea what % might be the margin of error? A lot of games have no winner reported - for example, in my head-to-head against my Terran friend between two of our accounts, it was something like I won 9 games and he won 1 game, but out of the 11 or whatever games with no win decided according to the site, he won most of them. If everyone has the same experience as we do, then Protoss would be heavily favored lol. Perhaps something to look into - if the undecided winner games favor/disfavor certain races. It's less likely the algorithm can detect winners in reps prior to 2009 as they contain less info. And as said above, if the winner is unknown, the game is not included in win ratio. Yes, I meant that certain match-ups are harder to calculate the winner for when it is one race over another, for example Protoss is easier to determine to have beaten Terran than vice versa, resulting in more PvT wins being reported and more TvP games being undecided, thus influencing MU win%. The winner detection algorithm does not depend on the races.
You said that one of the things it determines "largest remaining team", how is that calculated? Couldn't it have some racial factors? 😅 Sorry if I am not understanding, just working off of one anecdotal example.
|
Dakota_Fanning
Hungary2371 Posts
On April 20 2021 15:32 Dakota_Fanning wrote:Show nested quote +On April 20 2021 09:40 zimp wrote:On April 19 2021 22:19 Dakota_Fanning wrote:On April 18 2021 15:50 Dakota_Fanning wrote:On April 18 2021 06:12 zimp wrote:
Congrats on your database!
To contribute, i uploaded 50k replays, out of which only about 15k were not duplicates. Still something.
"...replays do not store the information who won. I wrote some algorithms to detect winners based on what’s available in replays, but it never will be 100% accurate."
I did not find race winning stats in the global stats page. Could you please extract tvp, tvz, pvz winning stats overall, by year, and by main maps?
Thank you!
A few more development ideas:
- in the player search, there could be an option for "player name includes", becaues the clan tag is often before the name of the player, and this way the results don't show up.
- more than 20 hits could show up on one page, or the user should be able to modify the default.
- if there are many pages of hits, it shoud show how many.
- there should be an option to "select all" and "download all" Very nice, thank you. Overall race / matchup stats do not yet show winning stats, I will add those. All maps already include matchup stats. Look at the map profile, e.g. here's Fighting Spirit, you can see overall race stats and also broken down for each matchup played on the map. About player search: there is already a "Anywhere in players" search type, that is equivalent to "player name includes" Yes, I plan to customize the page size. Showing how many results there are requires running the query twice, increasing the server load and response time by 2. I also plan to add support for downloading multiple reps. OK, global race stats have been added, by matchups. Visit the Matchups page to see win ratio of all non-mirror, 2-team balanced matchups. From that: P > T > Z P > Z (53%) P > T (57%) T > Z (54%) (Obviously this is just calculated from the uploaded 1v1 replays.) Thank you! That is impressive! This is the largest dataset from which race stats were calculated (that i know of). Now unfortunately that confirms my guess, namely that starcraft bw is not a balanced game, if we consider all levels. It is skewed towards P. Would it be possible to add a year filter? Or just have stats for each year somehow. To see if the evolution of the maps / strats actually changed the balance? What algorithms do you use to determine the winner, and do you have any idea what % might be the margin of error? Currently it's not possible to show matchup stats by year as it is not gathered separately. Might add it in the future. The algorithm I use is rather complex, but essence of it is tracking who leaves first, and the "largest remaining team" wins. This tracking is less accurate in replays before 2009, as those replays contain less information the algorithm uses. If anyone's interested, I open sourced the replay parsing engine including the winner detection algorithm here: https://github.com/icza/screp. I have no information about the error margin. But if the algorithm cannot decide who's the winner, that replay is added to the "unknown" part of the W-L-U record, and it is not included in the ratio calculation.
OK, it took me this long. We've arrived to the "future" mentioned above.
I've updated the Global Stats page to include Global "Everything" stats such as:
- Global Games Summary
- Global Race Play Stats
- Global 1v1 Matchup Stats
- Global Format Stats
- Global Type Stats
Also added Global Race Win % Trends and 1v1 Matchup Win % Trends data to the Year Stats table. Enjoy
The Global Race Play stats shows (this includes all types and all formats):
P win rate: 54%
T win rate: 49%
Z win rate: 50%
Global 1v1 Matchup Stats:
PvZ: 53%
PvT: 57%
TvZ: 54%
|
Ok. Thank you for adding these!
So if i understand correctly, in case of 1v1 reps, the winner is based on who leaves first/last.
We can safely assume that for the majority of the games, this will be correct. There will be many cases when this will not be correct, but this will not effect the race stats, as you can expect this behaviour from any race (that the person leaves before winning), and with large numbers, the occurrence should be balanced.
![[image loading]](https://i.imgur.com/62e4Hb3.png)
Based on ~600k games, again, this is the largest dataset that i know of, although it's an overall of 20 years. I don't really understand the peak in 2009.
It seems that after 2011, we see an increase in the imbalance. I simply don't know what causes this. It doesn't seem to be Python and FS.
I tried to match these stats with the other 2 large dataset that I have (400k from Iccup 2009 and 180k from 2016 from another source), but they don't seem to correspond in the percentages.
![[image loading]](https://i.imgur.com/0HWh6Xm.png)
In case you are interested, here are also some pro stats:
![[image loading]](https://i.imgur.com/1snhhLW.png)
|
Dakota_Fanning
Hungary2371 Posts
On April 21 2021 02:39 zimp wrote:Ok. Thank you for adding these! So if i understand correctly, in case of 1v1 reps, the winner is based on who leaves first/last. We can safely assume that for the majority of the games, this will be correct. There will be many cases when this will not be correct, but this will not effect the race stats, as you can expect this behaviour from any race (that the person leaves before winning), and with large numbers, the occurrence should be balanced. Based on ~600k games, again, this is the largest dataset that i know of, although it's an overall of 20 years. I don't really understand the peak in 2009. It seems that after 2011, we see an increase in the imbalance. I simply don't know what causes this. It doesn't seem to be Python and FS. I tried to match these stats with the other 2 large dataset that I have (400k from Iccup 2009 and 180k from 2016 from another source), but they don't seem to correspond in the percentages. In case you are interested, here are also some pro stats:
The main difference is the input dataset imo. Repmastered.app has multilpe hundreds of thousands of fastest games (big part of which are 1v1 games), and your other tables only contain low money maps. So you'll get closer results by comparing 1v1 matchup stats of specific maps, e.g. | iCCup | Fighting Spirit and | iCCup | Python 1.3.
|
|
|
|
|
|
EPT
