As we are on the last day of the season prior to the Oct 6th reset, I figured I'd take a quick look at the last two weeks as it was clear it performed very differently to that of season 1.

The first graph shows the normalised peaks for the last two weeks of both seasons, they produce the same graph as showing just the peaks, but it helps to see the variation. it should also be noted that the vertical axis starts at 70% to further emphasis the differences.

It is important to understand that slopes here do not necessarily equate to the degree of say player loss. So one needs to cross-reference against the second image/graph that shows the actual daily peaks for both weeks in both seasons. For completeness i've also added today's steamDB results as the third image.

At the start of the fortnight we see that the penultimate Tuesday update has very similar performance for both seasons. However come the actual weekend S2 maintains engagement far more, before having a significant drop-off, whilst S1 was far more gradual and never fell quite as far. This drop-off continues until the actual final weekend, when engagement rises significantly verses S1.

This raises the point that whilst initially well received the final Tuesday update of season 2 did not retain players into the start of the second week as it did in S1. Although this needs to be considered with respect to the change we've seen during the end of season 2 towards more engagement at weekends.

Specifically what I think we are seeing here has been a progressive movement in the last month or so of S2 towards players becoming more focused into the weekend and less during the week (with exception of well received Tuesday updates). This is likely encouraged by for example Cryo opening times, but I'd argue its clear that S2 changes to the game, its modes and options has added to this focusing of the player base into the weekend.

There is likely more to it than just this. For example I should have included a week from earlier in the season to demonstrate how the engagement across a week has altered during S2 and re-examined exactly what Bungie did differently this time, but sadly I do not have the spare time at present.

Below is the data used.

Unsurprisingly Fri/Sat peaks continue to show positive signs with the 'end of season' activities (Double XP and Sponsored Survival).

For the third week in a row the Fri/Sat peak has exceeded the Thu/Fri 'Cryo' peak. Which is interesting as the actual week-on-week was slightly less than Thu/Fri (18.5% vs 19.9%) and the difference in the number of players between the two days was half the week before. This may indicate that Fri/Sat peak is falling back and next week may be less than Thu/Fri.

Additionally for the last two weeks Sun/Mon has taken the crown for the lowest peak of the week, which has been traditionally Wed/Thu. Though looking further back Sun/Mon has occasionally been the lowest peak of the week so it might be worth looking into those weeks sometime to see if there was a reason for that.

Looking at daily week-on-week numbers this weeks Tues Update posted a huge 34% gain on the week before, though that has dropped each day since to todays 18.5%. If this trend continues it would strongly suggest that the game has a constant decline of player peaks baked in. That even when given extra incentive to play the current player base is unable to staberlise. Will have to keep an eye on this claim next week and see if it holds true.

^(- Number in parentheses in the title is the week index of the year. Marathon released in Week 10.)
^(- Days are given as Tue/Wed as the EU & US peaks straddle UTC midnight. So Tue/Wed would be Tues night.)
^(- L-Drop: Drop from Launch Peak S-Drop: Drop from Season/Event Peak)
^(- Week-on-Week update for Fri/Sat [ approx peak - Sat 26th Sept 03:00 UTC])
^(- Player Counts come from) ^(https://steamdb.info/)
^(- Reported numbers should be accurate and correct, please comment if you spot an error.)
^(- If you comment that charts are sad, then you are a big poopy head.)
^(- Speculation and predictions are not guaranteed to be correct.)
^(- Graphs are provided for fun.)
^(- Week 38 Sat 19th Sept -) ^(https://www.reddit.com/r/MarathonPlayerCount/s/Q14FN8r1ii)

Marathon continues to surprise with a gain of 368 players over yesterdays disappointing Cryo results. It still saw a 2.6% relative drop week-on-week, which is what one might have expected towards the end of the season had Bungie provided an update to encourage players back, but it is bucking the trend of the previous week.

While it has bucked the trends, it is coincidentally following another trend where Thu/Fri vs Fri/Sat swap places for which has the highest peak. This is now the sixth week where they have swapped places, seven weeks in total. I've mentioned this previously, but i'm not sure if there is really anything that can be taken away from this information.

This result would seem to put pay to the claims that WARDOGS would affect the game. Maybe there was some loss yesterday, although even then the 16% relative drop was not unusual, but todays peaks would appear to be unaffected?

The EU peak was higher than yesterday, but it was really the US timezone that planted its flag, pushing the peak way above what would have been expected, though there is little that I'm currently aware of that would explain it. So for the time being just have to put it down to the churn we'd likely see from, day-to-day differences, the end of a season and where the low player counts mean small changes can lead to more dramatic swings.

Looking ahead this makes any prediction difficult. Is this a one-off or an indication that players are returning to the end of the season regardless of any Bungie updates (or lack of). The following week could really go either way, but I would postulate is heavily dependant upon Bungie's actions.

A good Tuesday update and news about S3 would in theory provide a bounce to the last week of the season, but even no action might mean the numbers remain static. However given Bungie's recent activity, I'm not convinced that Tuesday's update is going to be anything of note. It would seem a bit late to make a change to promote the end of season, and certainly too late in the season to make any major changes.

^(- L-Drop: Drop from Launch Peak S-Drop: Drop from Season/Event Peak)
^(- Week-on-Week update for Fri/Sat [ approx peak - Sat 12th Sept 03:00 UTC])
^(- Player Counts come from) ^(https://steamdb.info/)
^(- Reported numbers should be accurate and correct.)
^(- If you comment that charts are sad, then you are a big poopy head.)
^(- Speculation and predictions are not guaranteed to be correct.)
^(- Graphs are provided for fun.)

Shocking week for Marathon's player count as it continues to tumble, falling below 3k on Cryo Thursday is not something I would have predicted a few weeks ago.

At a 16.7% relative drop from last week it wipes out the 'holds' of the previous two weeks. Whilst that is somewhat normal for Thu/Fri week-on-week it is surprising towards the end of the season. Well at least in comparison to the end of S1. The biggest differences are the lack of a Tuesday update and the lack of any tweaks to encourage players to engage with the end of the season.

If the trends continue and this wasn't an outlier, then Fri/Sat is like to be around 2770, though the range could be as worrying low as 2680 - 2790. That lower value is an absolute worst case but is already getting dangerously close to 2.5k line! If things don't change there may be a real chance of hitting the 2.5k on Sun/Mon!

I'm not counting Bungie out yet, they still have Tuesday update to entice some players back for the end of the season and avoid dropping below the 2.5k. It wouldn't even surprise me for an inspired update to pull the counts back towards 3k. They might also drop the 'most amazing promotion for a season' ever ...

... or they do nothing.

EDIT: Didn't realise that WARDOGS released Thu/Fri and hit a peak of 340k Sure they are different games, but you've got to think that some proportion of the player base has some interest, so a good possibility that things may track even worse than I thought for this weekend.

^(- L-Drop: Drop from Launch Peak S-Drop: Drop from Season/Event Peak)
^(- Week-on-Week update for Thu/Fri 11th Sept [Fri 03:00 UTC approx])
^(- If you comment that charts are sad, then you are a big poopy head.)
^(- Player Counts come from) ^(https://steamdb.info/)
^(- Reported numbers should be accurate and correct.)
^(- Speculation and predictions are not guaranteed to be correct.)
^(- Graphs are provided for fun.)

Might post these graphs once a month to provide some additional context to the player count since the games launch.

It should be noted this is only showing each weeks Fri/Sat peak (Sat 03:00 UTC), so it will miss peaks & lows from other days, such as Seasonal launches, Vault Breakers etc. Not sure if there is a better way to show this on a weekly basis. I could just take an average or the weekly peak, but I don't think that will capture the same information. So currently its useful for just comparing the Saturday morning peak from week to week.

First graph shows everything, player count, percentage of peak vs launch, absolute and relative weekly drops, the subsequent graphs focus on just one of those data points.

I think the relative week-on-week change (note to self - should rename drop to change) is interesting in terms of just how volatile it is, though the player count seems to follow an exponential quite closely.

^(- Week-on-Week update for Fri/Sat [ approx peak - Sat 03:00 UTC])
^(- Player Counts come from) ^(https://steamdb.info/)
^(- Reported numbers should be accurate and correct.)
^(- Speculation and predictions are not guaranteed to be correct.)
^(- If you comment that charts are sad, without even reading the post, then you are a big poopy head.)
^(- Graphs are provided for fun.)

The downward trend continues after last weeks unexpected small relative gains and its slightly worse than yesterday. Again its a new low for Fri/Sat peak at 3381 and now less than 4% of the launch peak.

Unless these drops slow down in the last two weeks we may yet see peak player count at the weekend drop to the 3k line. Though I don't see why Bungie wouldn't have some decent updates to pull in a few more players for the end of the season, and one assumes they will have to start their S3 promotion soon which should revitalize interest.

^(- L-Drop: Drop from Launch Peak S-Drop: Drop from Season/Event Peak)
^(- Week-on-Week update for Fri/Sat [ approx peak - Sat 5th Sept 03:00 UTC])
^(- Player Counts come from) ^(https://steamdb.info/)
^(- Reported numbers should be accurate and correct.)
^(- If you comment that charts are sad, then you are a big poopy head.)
^(- Speculation and predictions are not guaranteed to be correct.)
^(- Graphs are provided for fun.)

Well the positive gains of the last week have evaporated coming into this weekend. Player count is a new low for Thu/Fri and we've hit 4% of the games launch peak.

At a 6.3% relative drop to last Thu/Fri its not a massive drop, and still far better than 15%+ that we normally see. However I can't help but be disappointed that the hold/gain of much of last week has failed to continue.

Of note the last two Friday's of Season 1 (see second table image) had relative losses of 7%, so it will be interesting to see if we've just hit that level early, or if the next two weeks can post lower declines. Even more of a coincidence is that the lower relative loss have both happened on 'patch' version 9 (1.0.9.0 & 1.1.9.0).

I still expect next week to be better as with only two weeks before the end of season, one has to expect Bungie will drop some updates to encourage players back in. So with luck we might hover around 3.5k from now till the end of the season.

^(- L-Drop: Drop from Launch Peak S-Drop: Drop from Season/Event Peak)
^(- Week-on-Week update for Thu/Fri 4th Sept [Fri 03:00 UTC approx])
^(- If you comment that charts are sad, then you are a big poopy head.)
^(- Player Counts come from) ^(https://steamdb.info/)
^(- Reported numbers should be accurate and correct.)
^(- Speculation and predictions are not guaranteed to be correct.)
^(- Graphs are provided for fun.)

Marathon continues its positive uptick of the last few days and for the first time on a Fri/Sat (outside of a new season or VB) has naturally increased its player peak!

It is still unclear as to why, though the general consensus from the replies is that S3 leaks are most likely to have inspired more interest in the game, along with a fortunate coincidence of a 40% unofficial sale via various third party key stores (NewEgg, GMG).

The real test will come next week, and while completely speculative, if player counts remain flat, then we have perhaps found a floor, if we see a further gain, then perhaps the game is starting to find its place and is retaining more new players than it loses, while a loss would indicate a welcome but temporary bump.

Beyond next week and I feel we are too close to avoid the 'end of season' boost that previously slowed the player peak decline to almost nothing and this time may even gain players. A few good updates and we could realize further gains on top. It may be premature and overly optimistic, but I wouldn't rule out crossing back up to 4k line before season end at this point.

^(- L-Drop: Drop from Launch Peak S-Drop: Drop from Season/Event Peak)
^(- Week-on-Week update for Fri/Sat [ approx peak - Sat 29th Aug 03:00 UTC])
^(- Player Counts come from) ^(https://steamdb.info/)
^(- Reported numbers should be accurate and correct.)
^(- If you comment that charts are sad, then you are a big poopy head.)
^(- Speculation and predictions are not guaranteed to be correct.)
^(- Graphs are provided for fun.)

Marathon continues to surprise and for once its positive!

For the first time the week-on-week was on par with the previous week. Ok, not quite, in absolute terms it lost 0.04% (36 players) and 0.95% relative to Thu/Fri of last week. Still it is as close to actually posting an increase (outside of a new season or Vault Breakers) we ever had.

The only problem for me is I'm not entirely sure why. Tuesday's update managed to buck the trend and post a small increase over Monday for a change. That was assumed to be due to the new Vandal White Rabbit skin, or maybe the Season 3 leaks, possibly even small improvements to Firestorm, but none really explain the continued positive trend we've seen since, even accounting for Cryo opening up.

Could this be the mythical floor? It feels a bit early for the end of season 'bounce' so I'm not sure what else there is other than just more people came back to play this week?

Tomorrow will be interesting, as we are probably already looking at a small gain over last week and if it can maintain this over the next week, then we may well be looking at a pivotal moment for the game. In fact with if this can continue, even as small as it is, I think it would bode well for Season 3.

^(- L-Drop: Drop from Launch Peak S-Drop: Drop from Season/Event Peak)
^(- Week-on-Week update for Thu/Fri 28th Aug [Fri 03:00 UTC approx])
^(- If you comment that charts are sad, then you are a big poopy head.)
^(- Player Counts come from) ^(https://steamdb.info/)
^(- Reported numbers should be accurate and correct.)
^(- Speculation and predictions are not guaranteed to be correct.)
^(- Graphs are provided for fun.)

Just like that we are back to our favourite 20% week-on-week relative drop (approx) and approaching 4% of the launch peak.

This week has been all over the place, but in the end we're back to the traditional weekly 20% relative drops and even managed to dip below the lowest prediction. If the trend continues there is a good chance we could be at the 3k line this time next week (3100 estimate), and may even have dropped below on Wed/Thu which is usually the lowest point of the week!

There seems to be no stopping this train at the moment. Maybe we'll see some slow down with a good 'end of season' type update, though we are still over 4 weeks away from S3.

The prediction range turned out to be pretty good, especially for such an atypical week. While it wasn't exact, I feel it managed to provide an improved reference point than just using the last weeks counts. Alas the weirdness of this week will adversely affect the prediction for next week, so i'll have to find a way to adjust the weighting. I don't think that's unusual, certainly not for something like player counts, which while they show trends, can also have atypical weeks, especially around new seasons etc.

^(- L-Drop: Drop from Launch Peak S-Drop: Drop from Season/Event Peak)
^(- Week-on-Week update for Fri/Sat [ approx peak - Sat 22st Aug 03:00 UTC])
^(- Player Counts come from) ^(https://steamdb.info/)
^(- Reported numbers should be accurate and correct.)
^(- If you comment that charts are sad, then you are a big poopy head.)
^(- Speculation and predictions are not guaranteed to be correct.)
^(- Graphs are provided for fun.)

Once again Marathon manages to surprise and disappoint.

After a positive start to the week where Mon/Tues & Tue/Wed managed to match and slightly exceed the previous weeks counts, Wed/Thu showed an unexpected decline. The decline was only 10% compared to previous weeks of 16% & 18%, but it was disappointing considering the positive start of the week. Yet worse was to come, with the traditional Cryo bump failing to match the 'gain's from earlier in the week, with a loss of 625 players and a relative week-on-week drop of 14% that is barely any better than the two weeks prior.

Unless Fri/Sat shows an improvement (not impossible, Friday and Saturday can swap positions as to the highest point of the week) it seems most likely that Tuesday's update had no effect on the downward trend.

One highlight, for me at least, was the decision to post the experimental trend prediction. Initially I thought it would be pointless as I expected this week to be atypical but it was difficult to know exactly how that would express itself.

Additionally I hadn't had the time to try and verify the formula's being used against previous weeks, so it was, and still is, very much purely experimental and not to be taken as 'fact' or even any statistical significance - yet.

In the end though I feel it provided a great reference point to discuss the player counts this week as it clearly illustrated the unexpected changes more than simply comparing to last weeks numbers.

It should be noted the predicted range isn't guaranteed, individual numbers here are not relevant, its the overall trend across the week and how it interacts with the real player counts that is illuminating. Thus if its above the prediction but below last week numbers, we can see its doing better than expected, but not enough to gain players. If its within the range, we know its following the weekly downward decline (that only major updates like S2 or VB have addressed), while if its below the prediction then things went very bad!

Assumptions

• Mon/Tue gains likely due to players last chance to play VB prior to its removal.
• Tue/Wed gains likely due to checking out update - new skins, new mode.
• Wed/Thu loss likely due to loss of players from update as nothing to keep them.
• Thu/Fri loss likely just returning to the downward decline.

^(- L-Drop: Drop from Launch Peak S-Drop: Drop from Season/Event Peak)
^(- Week-on-Week update for Thu/Fri 21st Aug [Fri 03:00 UTC approx])
^(- Player Counts come from) ^(https://steamdb.info/)
^(- Reported numbers should be accurate and correct.)
^(- Speculation and predictions are not guaranteed to be correct.)
^(- Graphs are provided for fun.)

Well, I thought this week would be an outlier of sorts and go against recent trends, but this goes way beyond that, so much so that it feels prudent to make a most before the weekend.

Firstly, apparently out of nowhere, last night we hit an all time low of 3507 players! Maybe I missed something in the main sub, but I hadn't noticed anything major about the update to explain this. This leads me to wonder if its an external pressure that caused this sudden drop, a new game or some other event going on that drew players away?

It could be as simple as the main bump from Tues update was to get new skins and thus drops off the next day. I need to start graphing the games Top Sellers and DAU positions from steamDB as that might provide some clue.

Last night (Wed/Thu) was always going to be the lowest peak of the week, as far as the typical trends for the game. However until last night, we were pretty much matching last weeks player counts, a feat which on its own is very unusual (maybe never actually happened previously), though I had speculated was possible, depending upon the reception to the update. Had this trend continued we would have (and still might) been looking at the first ever 'neutral' week where player counts didn't drop week-on-week, but didn't post any gains either.

The graph shown is 'fixed' to a specific starting date (Sun 10th Aug 10am UTC) and showing a specific period (one week) where this weeks player counts (cyan line) gradually fill in from left to right for each day we advance. The orange line is last weeks counts, the blue area is a tentative prediction range, that is likely only valid for when the current week and previous weeks follow consistent trends.

• Sun/Mon - Followed the predictive trend.
• Mon/Tue - Bucked the trend, even matched previous weeks counts - suspect this was players getting in before Vault Breakers was removed from the game.
• Tue/Wed (Update) - Well received by EU, but not as much by US (which is not unusual, probably due to US players getting on earlier than normal) and still matching last weeks counts.
• Wed/Thur - Traditionally lowest point of the week, but suddenly drop down lower than last week which is odd.

I'd argue its too early to tell where we will go for the weekend. Tonight should show a high peak due to Cryo and the 'start of the weekend', though in terms of highest point of the week it swaps between Thu/Fri and Fri/Sat. The important aspect will be to see where the peak falls. If it matches last week then we are likely looking at the first 'neutral' week for the game, substantial, but not exactly the best success to shout about. If it falls short, then its another failed update, though might show the power of skins for increasing revenue. If it falls into the predicted range, then thats even worse as it means the update had no real effect at all.

Edit: Did some checking and week-on-week yesterdays drop was only 10% (likely due to the gains in the days earlier), where as the two weeks prior it was 16% and 18% respectively. So while it was a dire night for the overall Peak dropping so low for the first time, it wasn't, in my opinion, as bad a fall as it could have been.

Wanted to play with some prediction tracking, though probably chosen the wrong time to do so as I suspect this weeks player counts will not follow recent trends. Indeed its already defied the typical week-on-week drop for Mon/Tues as it has matched and even slightly improved on last weeks peak.

As such don't treat this as anything more than a little experiment for fun. While it might have some potential for tracking values for typical weeks, I haven't yet had the time to verify this. However as I think this week is going to be 'interesting' I wanted to put this out sooner than later. My plan is to update it over the weekend.

We can see that the first day Sun/Mon the cyan line for current player count was nicely within the predicted range, however Mon/Tues provided a surprise, not only is it not within the predicted range, it managed to match, and even exceed last weeks counts!

I have speculated that this might be due to Vault Breakers ending so people want to make the most of it before its removed. Could be due to the interest in the Tuesday update pulling in more players prior to its release. Alternatively could just be a random chance as we have a pretty low peak player count anyway a few extra hundred players can easily sway the numbers.

Will be interesting to see how the counts progress during the week to see if any of these speculations play out.

Graph Information

• Week 1 = Player Counts from last week - the actual dates on x axis.
• Week 2 = Player Counts from two weeks ago.
• Week 3 = Player Counts from three weeks ago.
• Ahead The Cyan line = Player Counts for the current week.
• Light Blue Area = Prediction, range based on testing various algorithms such as weight ratio, averages, and for different windows (how many previous weeks examined)

An Explanation

While historical data can provide accurate predictions in a stable, well understood context, a typical games player counts do not necessary fulfil those requirements. In Marathon's case there are certainly groups of many weeks that definitely show week over week trends that could be predictable, but the unpredictable nature of weekly updates as well as other factors can easily throw it off.

As such do not mistake the predictions as explicitly stating that a future week must be within those ranges. These should not be taken as actual predictions, but as a means to evaluate a week based on the expectations set by the predicted ranges. Its just another tool to examine the trends of the player counts over weekly periods and longer.

What I'm mostly interested in, assuming that many weeks are predictable, is what happens during the weeks when a Tuesday Update throws the prediction off. Will the predictions be off for the whole week, or if the update is not well received will player counts return to within the predicted range by the weekend.

This obviously excludes new seasons and major updates like vault Breakers, as they have a substantial affect on the player counts.