|
=========
Introduction:
=========
+ Show Spoiler +- At the dawn of SC the only thing that a player has, to know if he is doing good over time is the supply count, but no game timer yet (Stone Age).
- The term APM has been introduced.
- Off-line replay analysers/organizers (BWChart appeared).
- Some tools that calculates the EAPM accurately (BWHF)
- Launchers like Chaos/mca64 introduced some plug-ins that allow some analytical infos during replays.
- UMS maps can show some useful live infos during game, like the numbers of workers.
- BWAPI has been used to extract some built-in infos during game/replay.
====
Now:
====
Players only tend to measure performance by ICCUP level, SC:R MMR, or ELO, win-rate vs ?, etc..
But this not relevant when you want to compare the performance in 2 games, or in a new season of ICCUP, or when you come back to SC after some hibernation.
Some players doesn't like to play/practice a lot while they are clueless about some basic factors that rules SC (they just want the fastest way to improve), factors that even top players fail to perceive, I mean they can see if they are doing something wrong through game/replay, but they can't measure/count it in the long term, so they can break their previous record, I mean everyone will tell you those tips:- build workers non-stop the early 10m.
- try to keep your minerals below 500 the entire game.
- don't overlook idle workers for long time.
But can we measure that between 2 games/2 players accurately ?
These are 3 Performance Counters that I'm planning to implement in replay after some tests (some things still in-progress), so any player can measure his performance and compare it to others much easier than before, without having to focus on the different strategies/probabilities SC has or APM SPAM just because pros does:
===========
Workers Cut:
===========
This counter counts the seconds in which your Nexus/CC has been idle, if you have 2 bases the counter will be doubled, 3 -> tripled, etc.., Top gamers has this counter increase to about 2:30 the early 9m like Bisu, but it differs according to the number/timing of bases as I said. Low APM player like me can make this counter doesn't exceed 0:30 despite the number/timing of bases (the 1st secret factor of improvement).
=============
Minerals Above:
=============
Counts the total time of being above 250/500/750/1000, and when was the last time you've been above 750/100 (so you know when in your BO your Minerals accumulates).
===========
Idle Workers:
===========
What if you have a company and your workers are taking a break and you don't even know, this will make you lose without knowing the reason, so having a measure for that will reduce losses, if you care about superiority (that doesn't mean to ignore priority). So this counter displays the number of idle worker, and the total lost time (as with Workers Cut, this increase when there is more Idle Workers).
If you think knowing these counters could improve players performance, or if you've more thoughts about performance counters/measures that could be measured programmatically, let us know.
|
One thing that comes to mind as an exception to workers cut: if the mineral count is low while workers are cut, the worker cutting might be on purpose. The best worker cutting is when the supply is also blocked; the 10/10/10 terran opening build.
Will this analyzer work for SC:R or just 1.16?
|
A lot of people would also use "time supply blocked" (a la SC2) to measure macro ability, and it's a helpful yardstick, but it's subject to the same build-dependent properties. And I think that's ok, you won't be comparing "workers cut" in a game you open 10/10/10/ to a game you open CC first, you know that's a bad comparison, so you'll either choose two more similar games or you'll automatically deduct e.g. 30 seconds from the 10/10/10/ number
I think resources lost/resources killed is a really cool metric as well. Again, the exact numbers shouldn't be 1:1 for different matchups (e.g. late game PvT) but is useful to compare two similar scenarios.
A final suggestion could be map vision, what percentage of the map is visible. Obviously a good number for this would probably be ~50%, 50+% is pretty ludicrous, but could be a good indicator of who has more bases/map control
|
trying to stay under 500 minerals the entire game is counter productive. You will need to accumulate enough to fuel your production and since you aren't constantly hovering your gates (or other production facilities) you will get to 1k+ minerals.
You may get to 1500 minerals for each cycle of production and it's really no big deal.
This should probably get adjusted to the amount of bases and production facilities.
|
follow this guide and you too can be as good as moataz
User was warned for this post.
|
On August 07 2018 20:16 Navane wrote:
One thing that comes to mind as an exception to workers cut: if the mineral count is low while workers are cut, the worker cutting might be on purpose. The best worker cutting is when the supply is also blocked; the 10/10/10 terran opening build.
Will this analyzer work for SC:R or just 1.16?
For all races there is inevitable worker cut delay before building the 1st supply (I'm not sure about Z), for P, if you split workers perfectly so #6 worker starts @00:14, the worker cut counter will be 6-7s.
I personally doesn't prefer cutting workers on purpose for any reason in the early 9m.
No, it works only for 1.16 replays, so until SC:R supports BWAPI or stop forcing people to use a specific ver, don't lose hope, maybe if its not for you, then for your children.
|
On August 07 2018 22:58 yubo56 wrote:
A lot of people would also use "time supply blocked" (a la SC2) to measure macro ability, and it's a helpful yardstick, but it's subject to the same build-dependent properties. And I think that's ok, you won't be comparing "workers cut" in a game you open 10/10/10/ to a game you open CC first, you know that's a bad comparison, so you'll either choose two more similar games or you'll automatically deduct e.g. 30 seconds from the 10/10/10/ number
I think resources lost/resources killed is a really cool metric as well. Again, the exact numbers shouldn't be 1:1 for different matchups (e.g. late game PvT) but is useful to compare two similar scenarios.
A final suggestion could be map vision, what percentage of the map is visible. Obviously a good number for this would probably be ~50%, 50+% is pretty ludicrous, but could be a good indicator of who has more bases/map control
"time supply blocked" is not a consistent factor to measure, it differs to your need, if you're 80/85, you can build 2 Tanks but you're blocked from building 1 BC.
Yes you can compare your worker cut from 2 different builds (early/late expansion), it's not a bad comparison, your main goal is to have the lowest worker cut time possible, theoretically I can make it 7-8s due to the inevitable worker cut delay before building the 1st Pylon, but it goes to 30s or more during the early 9m, again the number/timing of bases doesn't matter, I think you meant "worker count" instead of "worker cut" ?
The CoachAI has a measure for:
- Displaying live score of produced/killed/lost units/buildings, to get the current state of players, this is about unit/building control (more close to micro). (same as the score screen you see at the end of game/replay)
- When selecting a unit/building in replay, displaying how many of this type has been killed during the game.
I don't think a counter/measure for map vision is necessary since it can easily seen by eye. and if its done it might be extra info that complicate things more.
|
On August 08 2018 02:28 LML wrote:
trying to stay under 500 minerals the entire game is counter productive. You will need to accumulate enough to fuel your production and since you aren't constantly hovering your gates (or other production facilities) you will get to 1k+ minerals.
You may get to 1500 minerals for each cycle of production and it's really no big deal.
This should probably get adjusted to the amount of bases and production facilities.
There is 4 levels to be measured: 250/500/750/1000, if you reached 1500 and stayed there for a while the counter will increase (actually if you go above 1000), it's a long term measure which is better than what some UMS maps used to do (like warning you when you're above 1000).
Do you suggest another counter for being above 1500 ?
|
On August 08 2018 02:48 Jealous wrote:
follow this guide and you too can be as good as moataz
User was warned for this post
|
On August 08 2018 05:56 Moataz wrote:Show nested quote +On August 08 2018 02:48 Jealous wrote:
follow this guide and you too can be as good as moataz User was warned for this post
I actually didn't say that because I hate him, I was just responding to him trying to tease me, I just found that smiley in the forum relevant for the situation, I hope there is version of that smiley with smiling face, so it doesn't appear like I'm angry or hate the person I'm taking to.
|
Just tell him (a.k.a. Mr. ILOVEKITTENS) that a kitten dies every time he's snarky.
Maybe also implement an 'idle production building' timer? (But exclude floating buildings)
In another RTS with modern development, OpenRA, I've found live ingame counters of total earned / total lost really useful. Sure there'll be racial differences, but its still a bassline metric of "I straight-up won simply because I produced way more than you".
Could you incorporate a screenshot so we can see how the info is presented on the screen, if its realtime? (Presumably looks similar to CoachAI).
The mineral floating cap should be adjusted to how many production buildings you have, because as said before, some production cycles can cost that much. Or when you're 200/200 and banking to remax.
Do replays preserve viewpoint movements? I know they preserve selection, but what's been the progress on BW first-person replays? I'm remembering back to old IntoTheMSL NaDa VODs using Penguin Plugin on Neo Forte. Good times.
I heard a rumor a while back that the way Korean BW pros would measure someone's speed wasn't so much APM, it was how many 'screens' they looked at i.e. how often their view was centered on a completely different part of the map e.g. Bisu viewing his corsair harass, production, natural, and army = 4 views.
I don't know how you'd implement this, but it'd be a measure of "do you scroll around a lot, or if you jump and efficiently switch between mental tasks." Better grade for jumping longer distances between views, better score for jumping more frequently.
|
'idle production building' timer, I'm not sure if that is necessary, since there is other factors that urge the player for non-stop production, like the timer for Mineral Above, but if you still find it useful I can make it.
![[image loading]](https://image.ibb.co/jWqFOz/sc.png)
If you've a formula that calculate the Minerals Above consistently/better, while the other factors changes (bases, workers count, prod buildings), tell me.
If you're taking about CanRep, I just contacted the programmer about an issue where the camera sometimes doesn't move during the replay, and he send me a link to an updated version. and it works fine for watching replays as FPV.
Measuring the multitask skills programmatically away from the EAPM term is not right, the discussion should be about what is considered effective and what is warming-up and whats useless.
|
These are cool ideas. I think it's most useful when comparing the same player performing the same build. A little less helpful, but still probably interesting when comparing different strategies. It's obvious that any strategy which breaks a rule of thumb will require you have different expectations.
One thing that might make this more useful is if you can either guess when two replays appear to have used a similar build and group them together, or if you allow some way of easily flagging replays for the intended build, like some syntax in the filename. Could be as simple as everything before an underscore in a filename represents the build name, and those replays should be compared to each other.
I think one measure that might be interesting is suspicious drops in apm vs the average, as that might indicate a point of confusion worth reviewing, or a type of micro the player finds difficult.
You do need to account for not counting up when the player is at 200/200 for most of these metrics.
Minerals and gas spending could probably be measured separately. If you can measure idle nexuses, I guess you could measure idle gateways and such as well.
I really don't think it's a problem that sometimes it's better for us to micro and let our minerals build up, or edge cases like that. But it might be interesting to keep track of win / loss and see when certain combinations of measurements correlated with one or the other.
One last thing that might be interesting, but pretty tedious to implement, is breakpoints like speedrunning streams have, but for builds. IE times you expect to get a building or the first of a unit or an upgrade, and how much you were off by. This could maybe be done by taking the explicitly intended build order from the file name, and just comparing times certain important structures were constructed, like the first nexus, the first stargate, the first +1 weapons etc. This last metric would be a good general sense of how well you were doing on everything else, because if this metric fails it matters a lot less that you were good at the other stuff. It also gives a lot of flexibility in that if one thing is slower, but everything else is faster, you can get some insight that the one thing doesn't need to be that fast.
|
BWChart has features for searching/organizing replays, but I guess you know, just in case someone doesn't:
![[image loading]](https://image.ibb.co/btN8vp/bwchart.png)
I guess I do need the counters on when the player is at 200/200, to test the player ability to quickly find a gap and break/remax rather than waiting for enemy to max too. So I think it would be better if a timer for being above 190/200 supply exists.
I don't understand what you mean by "Minerals and gas spending could probably be measured separately".
If I'm going to measure the time of idle prod buildings, I'm going to consider adding Robo for Protoss but what if he is planning not to use it much, and when game is not TvZ then considering Barracks for Terran won't be right.
CoachAI has a Build Order log, is that what you mean:
![[image loading]](https://image.ibb.co/fxCmgU/bo.png)
|
Looks like an interesting tool, especialy for low - mid level players, who cut probe production by a mistake, rather than by choice.
|
I would say there is a difference between the build order you intended to do, and the supply counts you actually built stuff at. Whether that's because you messed up or because the circumstances of the game forced you to adapt. I believe there is also a concept of the same general idea of a build you are trying to do, but different adjustments you might make to it which you would still want a way of grouping together.
I was not thinking of those features, but I do recall them now. Honestly haven't played SC since 1.18. Of course I could play 1.16 on iccup (if that is still going), but something about needing your friends to be on different versions of StarCraft just really ruined that ecosystem, and I'm not able to run anything after 1.18. Lord Martin Replay Browser actually had a really sophisticated tool for filtering replays based on what happened in them, and really acted like a database which BWChart never did. Unfortunately LMRB's domain name expired and it's no longer hosted (don't try to go there now, it's some suspicious weird site).
What's missing is an easy and simple way of aggregating and comparing data, which I thought was the intention of your thread. Like BW Chart tells you about one replay, but obviously we should be able to easily look at charts comparing groups of replays, or the last two games or whatever. The more tedious the process, the less anyone will feel like doing it. So yes, the build order log exists (even in ye olde BWChart), but it would be nice to compare it, or see comparisons on the fly in game. That way people could upload what they thought was a progamer level of execution, and you could see how you stack up as you play.
I don't know that it is really useful to track how long a player is about 190 or at 200, but you could try it. Most players do just blindly attack at 200 just because, it's probably a positive indication of skill if you sometimes don't because it might mean you're reacting to a situation. Of course you can let your money build up for a good reason too, but it's a little more obvious that you almost always want to be spending it.
I said minerals and gas could be measured separately because it is a very common mistake to have difficulty managing your gas, because you build assim too early. Too many minerals can mean you build it too late.
I don't think there's any need to worry about people misinterpreting the data. A person will know whether or not robo idle time was suspicious or not.
Anyway, I was thinking more in terms of generating charts like BW Chart, but I get the impression now you meant more this live gui. One nice thing about doing it just with replays after the fact is that it would probably be compatible with people on SCR, so you'd get more users.
I really loved looking at BW Chart after every game back in the day. But I'm just tossing out ideas, as I said I don't really play anymore. The idea of creating metrics for regular players to analyse their progress is really interesting though. When I used BW chart, I would look at things like where my APM spiked, whether I was using hotkeys consistently through the game, what my apm was like at some time vs my opponent's... It was damn cool, I miss it 
|
On August 10 2018 11:15 Chef wrote:
I would say there is a difference between the build order you intended to do, and the supply counts you actually built stuff at. Whether that's because you messed up or because the circumstances of the game forced you to adapt. I believe there is also a concept of the same general idea of a build you are trying to do, but different adjustments you might make to it which you would still want a way of grouping together.
CoachAI BO works like that:- Only the 2nd worker is logged for each player, to see his split skill.
- Only the 1st supply unit is logged, you don't want every Overlord created to be recorded !
- By default units (Zealots/Dragoons, etc..) logged until 7:00.
- By default the supply is logged until 40/?, You don't care on how much supply the players build his 4th base.
- "replayLogUnitsFor": 420, "replayLogSupplyFor": 40 in the config/json file can be used to change the 2 previous rules live.
On August 10 2018 11:15 Chef wrote:What's missing is an easy and simple way of aggregating and comparing data,
Sorry, but this tool isn't replay organizer or offline replay analyzer.
But ChaosLauncher allow you to run unlimited instances of the game, so you can open 4 instances in W-MODE for 4 Protoss(Rain, Bisu, Best, Stork) and compare their stats/timing (for example who is better in timing for specific build).
I still don't understand what do you mean by: minerals and gas could be measured separately.
FYI BWChart isn't accurate for calculating APM and some other things, I'm using BWHF for getting the EAPM.
|
On August 09 2018 20:52 Bonyth wrote:
Looks like an interesting tool, especialy for low - mid level players, who cut probe production by a mistake, rather than by choice.
I would like to know why you think top/advanced player wouldn't see these performance measures much helpful, I mean no one has reached the end of the perfection road, and what are the things that those top players think it would make them improve (is there is a secret, or its just playing/practicing a lot ?).
|
On August 09 2018 07:39 Moataz wrote:BWChart has features for searching/organizing replays, but I guess you know, just in case someone doesn't:
Is there a new version out for BWChart? I don't see a checkbox which allows me to filter the replays according to the build orders in BWChart 2.00C which I use. Can you post a screenshot with that filter option?
|
I downloaded 2.00C, but as you said. It seems to work only for 1.04C and above but not 2.00C, read more here.
|
|
|
|
|
|
EPT
