344 replies to this topic

[Z Paradox]

C337Skymaster, on 18 May 2020 - 04:49 AM, said:

Thank you for fixing it. I've actually been among those quietly wondering and hoping that MWO will be ported to Unreal Engine at some point (hoping, among other things, that it'll fix the invisible walls issue).

Fixing hitreg is more of a problem for me, lost my back CT when shot to front too many times....

[Anomalocaris]

ERescue, on 18 May 2020 - 03:59 AM, said:

Additionally, as the sample is rather small, is the appearance that teams lose more than they win a pure anomaly due to lack of sample size?

Well, first you have to understand that when we draw conclusions based upon statistics we cannot predict every result, only that certain results are likely to occur at a particular frequency. If the sample size is large enough then the predictions become more accurate. But that doesn't mean that outliers don't occur.

In the case of MWO, experience has generally suggested you need about 100 games in a season to eliminate the vagaries of the match maker and get an accurate picture of your true ranking. However, that was before the merge. Given that the matchmaker is now completely borked, it may take substantially longer. One player who is pretty good (mobajobg) recently posted that it took him 500 games to break a 1.0 WLR this season. He is a 98% player who generally puts up AMS in mid-300s. The last 6 months he's been averaging a WLR closer to 1.5. For it to take him 500 games to break 1.0 suggests that the matchmaking, particularly for solo players, is really, really screwed up right now.

[ERescue]

@Anomalocaris: Thank you. Let us focus on me for a bit more and why I am puzzled.

If assumption ERescue = 24th = most probable and actual data result ERescue = 24th, true 4/85, what is the probability of the assumption being true? I think the answer is very low.

Going further... If assumption ERescue = 19th to 24th = most probable and actual data result ERescue = 19th to 24th, true 30/85. This looks more likely.

Next step... Assumption ERescue = 13th to 24th = most probable and actual data result ERescue = 13th to 24th, true 65/85. This has a very high probability of being true.

If I did not miscalculate, my average position is about 16th... (1 338 / 85 is roughly 15,74).

I guess what I am trying to ask finally boils down to this: Is the ingame indicator so badly broken that it telling me that there are no weaker pilots than I am whatsoever, utterly false?

[Anomalocaris]

ERescue, on 18 May 2020 - 10:09 AM, said:

@Anomalocaris: Thank you. Let us focus on me for a bit more and why I am puzzled.

If assumption ERescue = 24th = most probable and actual data result ERescue = 24th, true 4/85, what is the probability of the assumption being true? I think the answer is very low.

Going further... If assumption ERescue = 19th to 24th = most probable and actual data result ERescue = 19th to 24th, true 30/85. This looks more likely.

Next step... Assumption ERescue = 13th to 24th = most probable and actual data result ERescue = 13th to 24th, true 65/85. This has a very high probability of being true.

If I did not miscalculate, my average position is about 16th... (1 338 / 85 is roughly 15,74).

I guess what I am trying to ask finally boils down to this: Is the ingame indicator so badly broken that it telling me that there are no weaker pilots than I am whatsoever, utterly false?

Well, if we look at Jarl's average for last season, it tells us that there are about 1000 active players consistently scoring lower than you, and about 10,000 consistently scoring better and 3000 scoring in your AMS bracket. If we were to assume that you are at the top of your score bracket, you should be averaging better than 4000 people and lower than 10000 people. That would put you about 17th on average in your matches. Again, that's assuming that you are at the top of your bracket.

Now, it may be that this season we are seeing an influx of new or old players that are substantially worse (because they are new or their skills have deteriorated). And we also don't know how the match score distribution will change because matchmaking is rather different than it used to be. You may also be improving as you play more matches which would tend to place you higher each succeeding season.

Jarl's is a tool, and it is most useful in tracking your own personal progress. It can also be used to judge relative progress, but it is less useful in that regards.

[ERescue]

@Anomalocaris: Appreciated, thank you! I guess I will look at things again once the month ends and both of my accounts get their latest stats.

[Janet Yellen]

Anomalocaris, on 17 May 2020 - 10:01 PM, said:

I think I agree with your post but did you mean average match score instead of PSR?

I would if PGI actually used average match score but they currently don't, and its seems that they have no intention to ever do so. Asking them to change two things without them $ing up more things is unrealistic. I'm going for the simplest change that can make a difference.

[Aidan Crenshaw]

What would that do, though? Your initial placement in the lances only determines where you spawn. Anyone can assume leadership at any time and switch lancemates around at will.
If they (PGI) wanted, they could the weight class distribution more strictly (as shown in the OP from Paul) and we would see more evenly distributed weight classes. At the cost of time-to-match, of course.

[Tarl Cabot]

ERescue, what is your average FPS?

Anyhow, here is one way of looking at things, and this is simply the average MS across X amount games. This person would have an overall average 180 matchscore. I broke it up a few different ways. The most high/lows could be even closer with fewer outliers.

Quote

Or another way to put it for avg Match Score

On wins, player avgs 300 MS across 1000 games, avg 100 MS across 1500 games and avg 180 MS across 3000 games. That would give a player an avg 180 MS.

Or 375 MS vs 50 MS / or 255 MS and 130 MS - this keeps the player medium rise both ways. Then a mix of circling around 180 MS mark.

• avg 297 MS - 2K games
  
• avg 150 MS - 4k games
  
• avg 105 MS - 1.5K games

That averages out 180.2 MS. Mix that up on win/loss column.. iie, I will let someone else try to calculate the other by giving PSR values and put that into a formula...


Remember though, avg 180 MS player in Tier 1 or 2 will likely have 25K+ games total. Of the above, the player was okay for 2K out of 5.5K games And damage generated, half of it counts towards the MS. So anywhere from 100 to 400 damage pts but likely being generous.

As for tier movement which uses PSR thresholds based on Matchscore, this is what it looks like. Not since its inception has PGI tried to directly fiddle with it even a little bit. How blind can they be?

Edited by Tarl Cabot, 20 May 2020 - 02:06 PM.

[Kyoptik]

[color=#959595]From an EU-timezone fan, I can only say thank you PGI, this is the first time group play with a friend of mine has been playable (<10 minute wait time) for over 6 months. I'd forgotten how much I love this game.[/color]

[color=#959595]Apart from queue times consistently under 2 mins, there's been less Nascar, more team play, more communication and a return to the awesome fun I had when I first came to Mechwarrior Online. Please please keep the changes![/color]

[Tarl Cabot]

And be mindful, I believe the current setup can be improved on.

• Remove tonnage limits - instead set to 1 per Class. This no longer the GROUP queue. This removes the potential of having 2 extra assaults, especially if competitive players. Again, appearances. And prevents the causal player group from bringing two lights. it would appear that many causal groups are not utilizing most of the available tonnage while the more competitive/effective groups are, and it would not matter if both types are Tier 1 players.
  
• MM can setup matching weight class for the solo players, with some regard to the odd pilot. 8 players vs 9 players.
  
• Reduce group size from 4 to 3. Effective groups are still a force multiiplier, Again, appearances and expectations
  
• While MM utilizes Tiers to make games, PSR needs a heavy revision, increasing the thresholds for movement and how much.
  
• Would be nice but doubtful - a Tier reseeding of the active player population based on average MS for the last 500 games, at least.

PGI has the license for a few more years. And without a major revision, they should be looking at how to even things out.

[DAEDALOS513]

Kyoptik, on 19 May 2020 - 03:27 PM, said:

[color=#959595]From an EU-timezone fan, I can only say thank you PGI, this is the first time group play with a friend of mine has been playable (<10 minute wait time) for over 6 months. I'd forgotten how much I love this game.[/color]

[color=#959595]Apart from queue times consistently under 2 mins, there's been less Nascar, more team play, more communication and a return to the awesome fun I had when I first came to Mechwarrior Online. Please please keep the changes![/color]

Considering all the positive feedback and population increase, they would be kicking themselves if they reverted it. What they have to do next is implement tweaks.. like some of the suggestion made above by Talbot.. I agree with all his suggestions except his first one about removing tonnage limits and going with class limits. The whole point of dropping with friends is so that you can actually PLAY TOGETHER. If one brings a light and the other an assault or heavy, you will not see your mate for most of the match and lance tactics would be thrown out the window (and with it much of the fun of dropping with a friend). Tonnage imbalances aside, I've been a part of many matches where tonnage imbalance DID NOT weigh favourably for the heavier side; it comes down to skill level. This is why it's important to set up effective psr system and mm.

Once population rises enough they could try splitting the queues again.. or adding scout mode?

Edited by DAEDALOS513, 19 May 2020 - 04:36 PM.

[Nearly Dead]

It is what it is. I think we can assume that PGI will continue the combined queue and make few or no changes to anything.

Guess we will see in a few months or a year where we stand.

[50 50]

Tarl Cabot, on 19 May 2020 - 03:18 PM, said:

ERescue, what is your average FPS?

Anyhow, here is one way of looking at things, and this is simply the average MS across X amount games. This person would have an overall average 180 matchscore. I broke it up a few different ways. The most high/lows could be even closer with fewer outliers.




As for tier movement which uses PSR thresholds based on Matchscore, this is what it looks like. Nor since its inception PGI has yet to fiddle with it even a little bit. How blind can they be?

You'd think trying a few different options for the calculation would be a good thing.
Remove the team win/loss difference.
Make the brackets for losing PSR the equivalent for gaining it and have a 'neutral no change zone'.
Given each tier bracket is a range of points we accumulate (the XP) then it would take a little while for players to filter between the tiers but should ultimately send players that are consistently getting high match scores upwards until they plateau and players who are consistently getting low match scores downwards until they also plateau.

[Anomalocaris]

Nearly Dead, on 19 May 2020 - 05:26 PM, said:

It is what it is. I think we can assume that PGI will continue the combined queue and make few or no changes to anything.

Guess we will see in a few months or a year where we stand.

You are almost assuredly right. At this point I'm just wishing I could wager on how big the decline will be over the next 6 months. Anyone selling short contracts on player growth? I'll probably just check the numbers periodically and come back for a final told you so in 6-8 months. So petty I know, but it's more than Cassandra got, no?

[Horseman]

ERescue, on 18 May 2020 - 03:59 AM, said:

There are multiple really poor games in that series and only a handful of really good ones. But my basic question is this: If I am statistically likely to be the weakest player out of all 24, how does the result level vary as much as above?

There will always be outlier matches where any given player dies too early or is ignored and racks up incredible amounts of damage and kills.
"Strength" of a player isn't defined as being able to high the highest damage or match score once, it's about being able to consistently hit higher values (and in the long term, higher averages).

Quote

Additionally, as the sample is rather small, is the appearance that teams lose more than they win a pure anomaly due to lack of sample size?

Teams will sway the results one way or the other depending on their competence (four comp players running meta builds will not perform the same as four casuals memeing around) but there is no guarantee of winning or losing, only a higher or lower probability of a given outcome.

Edited by Horseman, 20 May 2020 - 04:15 AM.

[ERescue]

@Horseman: You have been helpful in the past and now, but I will try to rephrase based on the recent results on the previous page in the post you quoted. The ingame indicator suggests that I am at the very bottom of tier 5. However, the average match results (from those 85 matches) suggest that I am approximately 16th out of 24 in the games I play. That does not feel consistent. And specifically, 81/85 I am NOT last, which does not feel like an outlier (or at least not my understanding of the word).

As for the topic teams. I do understand what you saying about different team qualities, but the question is rather this: If a pair of single PUGs is replaced by a pair of PUGs of identical quality, why does it appear (in the small sample) that two non-paired PUGs do better than the two paired PUGs?

For additional clarity (hopefully), it seems like 12 PUGs is superior to 2-man + 10 PUGs unless those two are of far higher skill than the other 22 people. And that feels really strange. As said, it is entirely possible the sample is simply too small (or there are more very unobvious 2-mans around, because I most certainly did not poll everyone in every game about their groupings.)

Additionally... I did not really write this down, but if my recall is correct, those 85 matches had a grand total of one with one 4-man group and even verifiable 3-mans were really rare. Again, given the smallish sample, this too may be pure coincidence.

[Anomalocaris]

Would just like to point out that at this point we're clearly seeing off-peak player numbers return to levels from before the merge, in some cases lower which means wait times during the periods when population numbers were the biggest issue should be returning to pre-merge levels barring any matchmaker loosening (peak play times never saw serious wait issues).

Peak player numbers are still up a 5-6% vs pre-merge. Furthermore average player numbers for the last 30 days continue to drop. We hit a peak of 520 avg players per day last weekend. We are now down to 512. Avg players per day in April were 466. At the present rate of decline I expect we'll be back below 500 within a week. Keep in mind that after 3 months of declines to start the year we saw the following jumps:

Feb - 308
March - 375
April - 446

https://steamcharts.com/app/342200#1m

[Horseman]

ERescue, on 20 May 2020 - 05:54 AM, said:

@Horseman: You have been helpful in the past and now, but I will try to rephrase based on the recent results on the previous page in the post you quoted. The ingame indicator suggests that I am at the very bottom of tier 5. However, the average match results (from those 85 matches) suggest that I am approximately 16th out of 24 in the games I play. That does not feel consistent. And specifically, 81/85 I am NOT last, which does not feel like an outlier (or at least not my understanding of the word).

The key thing to know is that Tiers are not rankings.
Your PSR change isn't based on your performance relative to other pilots.
If you consistently earn more than the minimum score to advance, you will advance until you max out Tier 1.
If you consistently don't earn more than that, your bar will either stagnate or degrade until you hit the bottom of Tier 5.

Edited by Horseman, 20 May 2020 - 11:21 AM.

[ERescue]

@Horseman: Alright, thank you. I will just disregard the ingame bar from now on.

[Brain Cancer]

Z Paradox, on 18 May 2020 - 06:12 AM, said:

Fixing hitreg is more of a problem for me, lost my back CT when shot to front too many times....

Some 'Mechs there's a small bit of "back" armor that can be hit from the front from high enough angles, and that can get you backshot from the "front" at times even without weird game mechanics.

More likely it's just the projectile ghosting through via a bit of lag just far enough that the game derps into registering it, though,