283 replies to this topic

[Nightbird]

thievingmagpi, on 02 July 2020 - 09:29 PM, said:

According to who exactly? You?

Or the people who the proposal was intended for, who apparently felt contrary to your own views, and are the ones who are going to actually implement the changes?

According the post which did not attempt to present benefits to match making or reducing stomp. Do you need a link?

[thievingmagpi]

Nightbird, on 02 July 2020 - 09:30 PM, said:

According the post which did not attempt to present benefits to match making or reducing stomp. Do you need a link?

And yet such an apparently dreadful proposal was adopted and your was....???

Lmao.

Swing and a miss.

The people that needed to be convinced of your system's supposed superiority were not.

That's your fault and your fault alone.

Edited by thievingmagpi, 02 July 2020 - 09:34 PM.

[Ekson Valdez]

Last warning. Stop the fighting or this will be shut down and everyone will enjoy some free-time off the forums.

[Nightbird]

Let's just wait and see whether the predictions pan out, shall we?

[Stormpaw]

Ekson Valdez, on 02 July 2020 - 09:33 PM, said:

[mod]Last warning. Stop the fighting or this will be shut down and everyone will enjoy some free-time off the forums.[/mod]

cmon, this s*** is funny

[Sjorpha]

So what I get out of this discussion so far is this:

Nightbirds system was in scope and had objectively better predicted results according to all simulations, no serious attempt has been made to argue this point in this thread so far.

The argument is rather about the fact that Nightbird failed to convince enough of his audience, no one really argues against this point either, least of all Nightbird himself. It's evidenced by the simple fact that it wasn't chosen.

So the real failure is the following: PGI and the community failed to base their decision on the facts, instead it was made based on a combination of uninformed preferences, politics and personal grudges, and a so one of the suboptimal solutions were chosen instead.

As far as responsibility for this failure, I consider it unfair to blame anyone of those who put real honest work in. So I'm not going to blame Nightbird or Jay Z etc. The real responsibility falls on pgi for not hiring or outsourcing the decision to a professional with relevant competence. Some guilt also falls on those who knowingly pushed for a solution they knew was bad, or who spread disinformation (such as the claim that Nightbirds model could not be implemented within given parameters).

Now also worth noting is that Nightbirds model, according to himself, is still not very good compared what could be achieved with an unrestricted redesign of the matchmaker. The demonstrations are of a principle that works based on his competence in math and statistical models, not of the best system possible. Undoubtedly there are professionals/specialists on the mathematics of matchmaking that could do better. No one has argued against this either, but I think it's worth pointing out.

The rational position is to argue for the best solution on offer, regardless of who made it.

[Natural Predator]

I came sum it up more easily for you:
Magpie: we polished up a **** and made it sparkle and used our influence to get it rammed through.
Nightbird: hey I have a diamond way to do this and we don’t have to polish turds anymore.
Magpie: your diamond sucks because our shiny **** was implemented
Mwo community: ?????

Edited by Trashtier, 03 July 2020 - 03:59 AM.

[Thorn Hallis]

You guys are bickering about a band aid being red instead of blue while the patient is bleeding white from dozen unaccounted wounds.

[Nightbird]

Thorn Hallis, on 03 July 2020 - 04:08 AM, said:

You guys are bickering about a band aid being red instead of blue while the patient is bleeding white from dozen unaccounted wounds.

Sorry, no band-aid here.

[Nightbird]

I scraped the leader-board data today and graphed it below. For those who want to reproduce the graphs, please note that when I scrape the data, I only plot pilots with >=20 games played. This is because when I created the sim, I discarded pilots with few matches as they can easily by chance have very high or low WLR.



The prediction WLR graphs. I'm thinking now that naming the prediction graphs at 25k, 50k games etc is not the best idea since we will not see the same number of pilots as in the sim. Instead, we should probably look at average games played (25k=60 games, 50k=120 games, etc) to line them up with the predictions.

We're 2/3rds the way to the first graph in the top left.



Lastly, here's season 47 with data cut before the patch. As you can see it's our current WLR graph is pretty rough right now but is predicted to get better.

Edited by Nightbird, 03 July 2020 - 09:53 PM.

[Tarl Cabot]

Will say this, both should chill.

The PSR setup was selected based on Like votes on that post. The bad part of that was that the actual proposals, especially the late comers to posting it on the thread then got missed or overlooked because said post did not show up on the 1st or 2nd page, NOR did PGI have said posts moved to the front.

Nightbird post showed up in that thead on page 13 and only had 11 votes while the post that bring the current PSR setup to chosen with 135 votes was on the first page, right after Paul's initial post. And your post gave a basic setup but readers had to be redirected to another thread to read the details, ie your post under Paul's thread, the presentation likely failed due to that.

In the end though, the real issue is how PGI failed to manage the "community" input to ensure all actual presentation was viewable with a range of options.

I do have a question, from your initial post in your own thread, and even this one and the post in Paul's thread with links to other threads, there are posts with graphs but no actual formula, no excel sheet. Am I missing something? With an existing excel sheet then Paul could get a better picture of how something could look without having to try to recreate it, either by himself or a coworker who enjoys formulas, excel sheets, etcs..yes?

And if there is a link to an excel sheet and it is not in the first post, then would that not potentially caused the presentation to fail in this type of environment?

[Nightbird]

Tarl Cabot, on 04 July 2020 - 07:07 PM, said:

And if there is a link to an excel sheet and it is not in the first post, then would that not potentially caused the presentation to fail in this type of environment?

WLR doesn't require a convoluted formula, being simply wins/losses. I'm not sure how a spreadsheet to show what happens after a match is needed? It's why I prioritized showing that it is effective in making better matches.

On the other hand Jay Z's proposal was only about the complicated formula. It may have won, but from my point of view it did not win from the merits of it's presentation, having not attempted to present evidence it would do anything. Though, had it tried, the authors would have realized it did nothing, and thus not submitted the proposal in the first place.

I didn't treat it as a popularity contest. Jay Z's group didn't treat it as a scientific contest. Which of the two should it have been? If the community treated it as a popularity contest, and as a result picked an ultimately damaging update, does the fault not fall with the authors and the community for not doing a rigorous job? I thought it would be entertaining to point out it really has been a case of the kettle calling the pot black.

[VonBruinwald]

Nightbird, on 04 July 2020 - 09:02 PM, said:

On the other hand Jay Z's proposal was only about the complicated formula. It may have won, but from my point of view it did not win from the merits of it's presentation, having not attempted to present evidence it would do anything. Though, had it tried, the authors would have realized it did nothing, and thus not submitted the proposal in the first place.

Jay Z did good making his system, but even he acknowledged the flaws, the problem was the lobbyists who pushed it. I'm still trying to decide whether they did so for popularity reasons or because they saw the flaws and their eyes lit up seeing how they could abuse it for their ePeens sake.

[Nightbird]

VonBruinwald, on 05 July 2020 - 12:35 AM, said:

I'm still trying to decide whether they did so for popularity reasons or because they saw the flaws and their eyes lit up seeing how they could abuse it for their ePeens sake.

I'm sure they genuinely wanted a better system to improve the MM, but the road to hell is paved with good intentions. There's no replacement for simple stupid math.

VonBruinwald, on 05 July 2020 - 12:35 AM, said:

Jay Z did good making his system, but even he acknowledged the flaws, the problem was the lobbyists who pushed it.

He saw what was happening, and always had the choice of ending the discussion by withdrawing his suggestion. It's what happens between professionals to keep the masses from getting involved. I think he believed his system would work, and also he also discredited my proposal by saying it was out of scope.

Edited by Nightbird, 05 July 2020 - 08:27 AM.

[Nightbird]

As of today:

[Warning incoming Humble Dexterer]

The old system just stacked everyone into Tier 1 over time, so things only got worse and worse as everyone reached Tier 1.

The new system stacked everyone into Tier 3, then splits them between Tier 5 and Tier 1 over time, so things will only get better and better as everyone leaves Tier 3.

The only issue right now is that there's too many players in Tier 3 : Something like the next loot bag event will kickstart getting players moving out of Tier 3, and only then can you judge the results of the PSR patch.

As for now now everyone in Tier 3 is even worse then having almost everyone in Tier 1.

[Nightbird]

Warning incoming Humble Dexterer, on 06 July 2020 - 01:02 PM, said:

The old system just stacked everyone into Tier 1 over time, so things only got worse and worse as everyone reached Tier 1.

Who is "everyone"?

[OneTeamPlayer]

Nightbird, on 06 July 2020 - 01:24 PM, said:

Who is "everyone"?

Aside, but with them admitting we struggle to make 1K weekdays and 1.5K weekends the only way those charts are correct is if they also place people who have barely played, as in everyone who has played 10 matches in 6 months or something silly like that.

The idea that we have 47 thousand-ish regular players is ridiculous on its face.

I'd be impressed if it were actually a quarter of that number.

[Tarl Cabot]

Nightbird, on 06 July 2020 - 01:24 PM, said:

Who is "everyone"?

OneTeamPlayer, on 06 July 2020 - 02:22 PM, said:

Aside, but with them admitting we struggle to make 1K weekdays and 1.5K weekends the only way those charts are correct is if they also place people who have barely played, as in everyone who has played 10 matches in 6 months or something silly like that.

The idea that we have 47 thousand-ish regular players is ridiculous on its face.

I'd be impressed if it were actually a quarter of that number.

The only fact Paul placed on that chart was that it was the number of players in each tier who had dropped in the last 2 months. We can only see overall numbers for the Leadership board if the player had dropped at least 10 matches.

May had 2182 new players/accounts while June had 1723 and April had 1650 new players. There are a ton of accounts which have not even played at least 500 games, even over several months time.

As of May 2020 Jarls have noted 27,958 active, players who have played and hit the Leadership board in the last 3 months. How does Paul's graph match up against those numbers? One is counted if they had played a game over the previous 2 months while the other records if they played during a season (month) to hit the Leadership board to even be publicly recorded.

Now what would be interesting is the ranges of avg MS per Tier as well as the range of games played over the lifetime of each account.

[Hellbringer]

Can a tier 4 or tier 5 player please step up and give your perspective on this new system. I’m truly interested in how your games are proceeding.