OSCORP: HEROIC Meta Analysis of Progress-Based Subs

I started a tracking thread here to gather data from the City sub of the latest run of Oscorp: Heroic. At the end of the City sub, aaronschmiz started a new tracking thread for the Tower sub here, and after crunching the numbers we collected from the City sub, I posted an analysis of the completion rate data in that thread. With this new type of sub, we aren't given a hard end time, and since timing refreshes is absolutely mandatory to maximize points, it's pretty important that we can start generating reliable estimates.

My post-City analysis is in the spoiler block below:
Vairelome wrote:
Analysis from City sub: I stuck all the numbers from the original tracking thread in a spreadsheet, and beat the data until a chart popped out, and found a few surprising details.

Expected:

The tail of the graph representing the final chunk of the sub did have a somewhat greater slope than the rest of the graph (though a good chunk of our early data was sporadic).

Not Expected:

The slope increase was smaller than predicted--midway through the event, we were accumulating about 1.5-1.7% every hour, but in the final period, gains were closer to 2.5% per hour. That's more, but it's not even double the rate.

The slope during the final period did not go exponential; while the rate did increase from the middle of the event to the end of the event, the final period was still basically linear.

Most Surprising:

So, the rate across the "final period" was about 2.5% per hour? That period started at around 80%, not 90% or 95%. We went from 79% to 90% in roughly 4 hours 30 minutes, and then from 90% to 100% in 4 hours 10 minutes.


Conclusion--I expect the data gathered from this first sub will probably do a good job of predicting future subs of this type, so long as the back-end mechanics stay more or less the same. Sure, regular forum-goers will be able to tune their performances a bit better, but I think this will have a minimal impact on worldwide trends, since we're a small minority of the total playerbase. Most players will check in periodically, grind some nodes, wait for refresh, see that the cumulative total probably won't permit an additional refresh, and grind extra during the last 80%, and that behavior will dominate the statistics.

I think that analysis largely holds up, though there are a few points that need clarifying. But first, here's a quick-and-dirty summary of the Tower sub rate data (all times will be east coast US; M/hr = million points per hour):

The first 15% of the progress bar took about 14.5 hours, from 6PM to 8:30AM, and the progress rate varied from 3-5M/hr. 15-30% took about 10.5 hours, from 8:30AM to 6PM, and the progress rate was around 6-6.5M/hr. 30-45% took 13.5 hours, from 6PM to 7:30AM, with a progress rate of about 5M/hr. 45-75% took 20 hours, with an average rate of 6-6.5M/hr. There was a brief dip in progress rate over the next 5%, down to 4.5M/hr, from 3:30AM to 6AM, but the next 15%, from 80-95%, picked up substantially, varying from 6.5-8M/hr from 6AM to 3:30PM. Finally, the last 5% was turbocharged, clipping along at 11-12M/hr from 3:30PM to the end of the sub at 5:13PM.

(In my City analysis, I used % per hour, though I think M/hr will be more useful in the future. For a completion bar of 400M, 1% = 4M.)

For the given set of variables that applied in Oscorp: Heroic, much of the middle phase of stacking up points happened at around 6-6.5M/hr. There were some dips in progress rate, almost certainly due to the sinusoidal nature of global activity across a 24 hour period, but even those dips had a floor of around 3M/hr early on, and 4.5M/hr later. The high-activity period at the end of the Tower sub was much shorter than the City sub, but only a little faster--City progress rate peaked at 10M/hr across the last 20%, while Tower progress rate peaked at 11-12M/hr across the last 5%. There are a few possible factors here--maybe the playerbase got slightly better at timing the final refresh; maybe the City ending on a weekend changed the global activity pattern.

In general, for subs of this type, there are a few different factors that will affect the overall rate of completion, and therefore the projected end time of the sub. The first is global participating player population. More people getting points = faster completion. Next is the base points on the sub's pins, probably best captured by the average value of the top 3-5 pins, compared to the overall length of the progress bar. Finally, there is the balancing effect of rubberbanding and scaling. Rubberbanding provides an acceleration effect where growth is smaller at the beginning (nobody is far enough apart to get bonus points), but much bigger later on (most of the player population is getting bonus points). Personal and community scaling provide a deceleration effect where pins are easy to beat early on, encouraging more grinding, but much harder later, forcing players to pick their battles and hit fewer pins. There's a second "acceleration" effect due to players joining the subs midway through the completion period, as well. In Oscorp: Heroic, the acceleration effects of rubberbanding and late joiners outweighed the deceleration effect of scaling, but it would take considerably more information to tease out the details of that balance.

Thank you to all the people who contributed data points to this tracking project; I hope we'll be better able to get a handle on progress-bar subs in the future with fewer headaches.

Comments

  • hurcules
    hurcules Posts: 519
    Many thanks for producing this analysis, Vairelome, really appreciate it!

    Since you already have the data points tabulated (in spreadsheet, I presume), would you mind posting the cumulative graph (both % and absolute numbers)? Would be interested to see how they look.

    Cheers!
  • I used an OpenOffice spreadsheet to collect and work on the data, but I don't know how to insert a graph into a post. I can put the raw numbers into a table, though, so you can cut and paste it. All of the numbers are from the threads I linked in the top post.

    City Data:
    Time         	Value         	Percentage
    01:34:00 AM	124000000 	31.00%
    06:56:00 AM	151000000 	37.75%
    02:20:00 PM	203000000 	50.75%
    04:17:00 PM	216000000 	54.00%
    06:23:00 PM	230657231 	57.66%
    08:35:00 PM	243000000 	60.75%
    02:30:00 AM	281000000 	70.25%
    04:07:00 AM	290289356 	72.57%
    07:27:00 AM	307000000 	76.75%
    09:09:00 AM	318000000 	79.50%
    10:40:00 AM	332000000 	83.00%
    11:10:00 AM	337000000 	84.25%
    12:43:00 PM	351907470 	87.98%
    01:19:00 PM	357667623 	89.42%
    01:34:00 PM	360000000 	90.00%
    02:20:00 PM	367400000 	91.85%
    02:28:00 PM	368000000 	92.00%
    03:00:00 PM	373438015 	93.36%
    03:30:00 PM	377903550 	94.48%
    03:47:00 PM	380000000 	95.00%
    04:00:00 PM	383280158 	95.82%
    04:20:00 PM	386115085 	96.53%
    05:06:00 PM	393512827 	98.38%
    05:30:00 PM	397500000 	99.38%
    05:44:00 PM	400000000 	100.00%
    

    Tower Data:
    Time	Value	Percentage
    06:21:00 PM	1673778	0.42%
    07:21:00 PM	4962771	1.24%
    02:16:00 AM	39347824	9.84%
    03:45:00 AM	44600009	11.15%
    04:05:00 AM	45675120	11.42%
    04:55:00 AM	48318175	12.08%
    05:30:00 AM	50182935	12.55%
    06:50:00 AM	53740959	13.44%
    08:38:00 AM	59783302	14.95%
    10:55:00 AM	71344613	17.84%
    11:53:00 AM	76872358	19.22%
    12:03:00 PM	78255171	19.56%
    12:42:00 PM	82287287	20.57%
    01:08:00 PM	84938525	21.23%
    01:45:00 PM	89209602	22.30%
    01:47:00 PM	89274077	22.32%
    03:12:00 PM	98775606	24.69%
    04:03:00 PM	104344315	26.09%
    04:25:00 PM	106788176	26.70%
    05:15:00 PM	111959721	27.99%
    06:03:00 PM	117328772	29.33%
    07:24:00 PM	124693878	31.17%
    08:41:00 PM	131254174	32.81%
    09:34:00 PM	135691426	33.92%
    10:58:00 PM	142599392	35.65%
    11:58:00 PM	147291551	36.82%
    12:25:00 AM	149808844	37.45%
    01:16:00 AM	154373531	38.59%
    01:17:00 AM	154580229	38.65%
    02:00:00 AM	157662817	39.42%
    04:34:00 AM	166871443	41.72%
    06:31:00 AM	173886783	43.47%
    07:24:00 AM	178397509	44.60%
    10:12:00 AM	194980973	48.75%
    12:09:00 PM	208603844	52.15%
    01:51:00 PM	219617104	54.90%
    03:52:00 PM	231101053	57.78%
    05:00:00 PM	238124635	59.53%
    06:13:00 PM	242028876	60.51%
    07:17:00 PM	248669507	62.17%
    08:23:00 PM	255851268	63.96%
    09:14:00 PM	260960673	65.24%
    10:35:00 PM	270184701	67.55%
    01:11:00 AM	288305575	72.08%
    02:21:00 AM	295472402	73.87%
    03:29:00 AM	301339052	75.33%
    04:13:00 AM	304736139	76.18%
    05:30:00 AM	309966090	77.49%
    06:10:00 AM	313097924	78.27%
    07:40:00 AM	321728948	80.43%
    08:02:00 AM	324081556	81.02%
    10:05:00 AM	339080473	84.77%
    11:13:00 AM	347380674	86.85%
    11:57:00 AM	353708810	88.43%
    12:48:00 PM	359851251	89.96%
    01:42:00 PM	366613207	91.65%
    02:50:00 PM	375973553	93.99%
    03:31:00 PM	381500000	95.38%
    03:45:00 PM	384320000	96.08%
    04:38:00 PM	393732980	98.43%
    04:51:00 PM	396299981	99.07%
    05:13:00 PM	400000000	100.00%