Separate names with a comma.
Discussion in 'Priority System Discussion' started by heymonkey, May 10, 2019.
Placeholder for explanatory text to come later.
Let me put down some objectives for the system that I think are a concensus.
1. Priority should be calculated over a period of games.
2. Players who have played in less games in the period should get priortity
3. When players have played the same number of games, whoever played less recently should have priority
4. There should be a bonus for replacing into a game
5. New Players should have priority
Then there other things that need to be considered
6. Should players dying early receive a boost
7. What happens to players playing in more than 1 game in a signup period. i.e. dying early then replacing into a different game.
I think my system of weighting each period fulfills the first 3 criteria in a really nice way.
If you sort by priority and scroll down towards the bottom scores you can see a nice cascade. At the top with players who have played 3 or 4 games with early deaths or replacements it can get messy. And the are some flaws to buff out in terms of what numbers should be used for points 4,5 and 6. The most obvious flaw is that a replacement score of 0.5 has ties between Faddy and Lone_Prodigy despite playing in different sets of games, there is also a tie between Stuart444/Z Beat and TheChuggernaut. This is not ideal.
I've been out since Thursday with family so I have work to catch up on first, and then I'll be in here.
God, I lost the thread. I still have to upload things into the first post, but one thing I'd like to see change is the replacement bonus/point allocation. I think either we need to adjust the points or give someone a spot in the next game they sign up for. I don't like to see people put on replacements twice in a row. Being a replacement is work and right now the bonus isn't enough, I think.
Also we'll need to talk to Pedro about doing the coding (we probably just want the real sheet managed by one person, with a backup, unless we're going to maintain a formal changelog), but having a dummy sheet for testing is a good idea, maybe. What's your formula you're using? Is it what you posted back in review?
The calculation is the score for the phase multiplied by the weighting number at the top of the column. Then all 5 scores are summed together.
So the top number is the sign up phase (1 to 5, 1 is the most recent) then the bottom number is the weighting each phase gets.
So if someone played in all 5 games the score would be (1*1)+(1*0.95)+(1*0.85)+(1*0.7)+(1*0.5) = 4.0
If someone has a boost they would get a number less than 1 multiplied by the weighting.
As for coding... the sheet I have is about as good as can be done tbh.
To update it all you do is add a new column C and shift everything to the left. Then put in the score next to everyone's name.
Then a little tidy up. Adding the game title to C3 and deleting C2 and C5 to move the cells to the left. (delete the games that no longer count or just have the data on right)
All the formulae will update correctly. The sheet is set up for the formula to directly reference the columns C to G so old data is discarded when moved into column H.
On this particular point. It is difficult to get this right imo.
If there is a game where two people with high priority scores sign up and are both on the replacement list. The player with higher priority subs in.
For the next sign up phase both players are still carrying high scores and neither make the roster for the next game on priority score. Your idea would give the person who played in a game priority over the other person who never got to play at all.
These kind of things are a draw back with doing priority over a number of phases. That people will build up scores and will be at the bottom of the list for multiple sign up periods in a row. Kopite and wee are in this situation right not.
Instead of giving someone a pass to enter the next game due to replacing in, I think simply not counting priority for games you replace into would be the better method. It's closer to the spirit of what we initially wanted with the 'priority boost' and avoids reducing someone's priority for subbing into a game on D7 who only got to play for a phase or two (which is an excessively crappy situation in and of itself).
As a side benefit it would resolve Faddy's question of what do you do for someone that played one game, died, and then replaced into the second when we have multiple games running at once.
Yes, that is the problem there. It should still count as playing in some measure, but the current measure does not do enough, I think. Ideally, if there is some point number assigned, the two people in this scenario both adjust after the game, but the one who didn't still ends up with more priority.
Right now, if both players are at 3:
X replaces in
Y does not
After the game, assuming one game dropped off for both (putting them at 2, without counting the current), scores would be:
If a game doesn't rotate out for either, or for only one, that's just the system working as designed. Though I am in favor of introducing decay.
I'm thinking more about adjusting that to .2 or .3 instead of the .5 it essentially is. Sure, some people will replace in day 1 or day 2 and basically get a whole game for less, but the idea is to maintain the spirit of willingness to assist that keeps people on the replacement list. I do agree it's difficult to get this right. It's probably the point we agonized over the most with the early iterations of this system.
So your aren't thinking we should have two different scores for early replacements vs late replacements. And the score should be low enough to not ever lose out on priority to someone that has regularly played more games.
So are we settled on 3 scores only?
Standard = 1
Early Death = slightly less than 1
Replacement = slightly more than 0
Honestly using 0.99 for early death and 0.01 for replacement works out quite well. You can see it on the sheet under "messing around" tab, it produces nice cascades of people who have played in similar games.
For me that does adequately fulfill points 1-4 and 6.
Do we need to do anything more for new players? I had them scoring the same as a replacement. The idea is essentially that like replacing in we give brand new players the absolute minimum score so they more or less remain at the top of the priority list. (That is factored into the messing about sheet, e.g. Bronson has a score of 0.01)
I think that if make the replacement score minimal it is fine to just add them together, the worst possible score is 1.01
I think using .01 for subbing in is fine, it's essentially the same as I was suggesting but gives a slight tilt against them in a perfectly mirrored situation between them and another player which seems fair all told.
Need to think on only .99 for someone that died early. In the scenario of two players that have only been in one game each - should the player that died D1 of the most recent game lose out to someone that fully played in game 2? A player that died D1 in the most recent game could lose out in priority to someone that played fully in game 2 and subbed into game 3, 4, and 5 to show an even iffier outcome (which I know this scenario is unlikely but just presenting it for argument's sake).
I assume we're also saying the .99 and .01 are not impacted by the game multiplier (.95, .85, etc.)?
Yes they do get multiplied
I feel it is easier to track that way. Instead of independently tracking sub ins or early deaths it just goes in as a priority score for the round. And therefore will naturally fall out of the calculation.
I think so. I see the purpose of the early death bonus to be for ensuring that those players get a boost over other people in the same sign up period. To have it do more than that doesn't feel right to me.
I don’t think the current spreadsheet required manual entry of anything except the sub ins as opposed to having to enter the whole score but Pedro would be able to speak to that better. It pulled the data automatically from the game data spreadsheet iirc.
But then what of the scenario where a player can have played in a full game and subbed into three others winning out over a player that played in one game and died D1?
The full game would be less recent than the early death.
And we are currently working on the basis that subbing in should have very little effect on priority.
Maybe subbing in needs to count for more? But then that brings up situations where a player who subs in loses out to someone who was originally in the game.
Ooof, we have one game left in the test period so I really need to get back to this. I haven't had much free time. I do have two points to raise, though.
What if we just do a simple degrade whereby each game starts to go down by .2? Most recent game is 1, before that is .8, before that .6, then .4, then .2. Not complicated. We could pull a random sample of people and test it.
Also I'd like to propose a rule adjustment. Anyone who replaces in late (see saenima in LB3) should just get the priority boost without the full point for the game. So for sae in this case it would just be. -.5. It doesn't have to be that, but I think we should recognize that replacing in late like that is not really playing in the traditional sense because it's so hard to catch up. You're really ONLY doing it as community service. I'd say this could be for anyone who replaces from D5 onward?
@Natiko? @Pedro? @Faddy? Anyone else have ideas here?
I don't have any issue with providing additional incentives to replacements. My main hold up at this point is what to do with early death/sub in/late sub in values.
I don't know that the decay values themselves matter a ton as far as what specific ones we go with, they'll all accomplish the same thing. Maybe an early death is just a flat .75 multiplier on top of whatever their score for that game is. Maybe subbing in is a .25 multiplier on top of whatever their score for that game is. If you sub in late do you just multiply by 0 and then subtract a flat value of maybe whatever the lowest max game value is (so .2 in Monkey's proposal or .5 in Faddy's)? Or you could do the middle game if you want to be more aggressive in providing incentives for subbing in late (so .6 in Monkey's and .85 in Faddy's).
Oh hell, I wasn't thinking about those because those are manual entries. You're right, if we keep current numbers, we're looking at a 70% reduction for early deaths and a 50% reduction (for now) for serving as replacements, with late game replacements serving as special case. But we need to make sure Pedro doesn't have to manually redo the math on those every time he moves a game down.
What all is and isn't automated on the current sheet at this point? I don't recall anymore, I know there were a few iterations of it.
I don't know. That's a Pedro question.
I'm still here.
I'll get back up to speed tomorrow :-)
The depreciation system doesn't need to be as fancy as I proposed because there are a few other factors that need to be considered especially around replacements but I don't think a simple degrade works. The numbers I put down does ensure that in normal circumstances the main factor in determining priority is games played in the period. Where as with the 0.2 deductions there are situations where someone who has played 2 games has priority over someone who has only played 1. I think that doesn't achieve the main aims of system.
I think we should stick to trying to achieve the 5 things I set out in my opening post.
I think we can get 80% of the way there with the system I put out.
I don't think we can achieve point 4 while keeping the other pieces in place. But that doesn't mean the system is a failure, it just means the replacement situation needs its own solution. And that a way of dealing with replacements that breaks some of the other goals isn't the end of the world.
For replacements, how about a system where you get a score of 0.2-(0.1*replacement day). So if you switch in on day 1 you get a small drop in prioriry. Day 2 is neutral then after that you get a boost in priority.
I don’t know if we need to go that granular on replacement boost. Mainly still waiting to hear from @Pedro on what things he has to do by hand. I don’t want to end up making it even more work to maintain.
I agree with that, the replacement system doesn't need to be complex. That was just an idea.