Some scary math that may explain why SoE thinks it's working as intended

[Zolina]

Say that pre-nerf, there was a flat 5% chance of a skill-up from 200 to 300. Then the average total number of combines to get from 200 to 300 would be (1 / 0.05) * (300 - 200) = 20 * 100 = 2000.

Now suppose SoE decided to make a linear sliding scale from 200 to 300 with an average skill-up percentage of 4% (20% harder than before) and a skill-up chance at 300 of 1%. That'd work out to a 7% chance of skill up at 200, and at any given skill level x, a percent chance of skill up of (19 - (3x / 50))%.

The average total number of combines to get from 200 to 300 is now SUM[x=200,299]{1 / ((19 - (3x / 50)) / 100)} = 3200.

So the chance of skill-up only decreased by 20%, but the number of combines required to get from 200 to 300 increased by a whopping 60%. We're getting a double whammy, first from the drop from 5% to 4%, second from the sliding scale, and the effect from the sliding scale is worse than the 5% to 4% nerf. As I see it, there was NO reason to drop the average skill-up chance from 5% to 4%, the sliding scale was more than enough deterrent.

[Phantron]

It is said that 250-300 is 20% harder, not 200-300. Your math is wrong. If it takes 20% more combines then it takes 1200 combines (on average) no matter how it's distributed.

Your model would actually have 250 starting at 4% and declining to 1% at 300, or 2.5% chance of skillup overall, which is twice as hard as what it used to be.

[Ngreth Thergn]

he brought it in from a guess I was making which was bad math

but I did indicate it would start at 5% at 250 and got o 1% at 300 so it would still take some ajustment, and call for the curve to maybe get steaper at the end. (not a straight line but a curve)

[Zolina]

I'll run the numbers again Ngreth, assuming 5% chance at 250 and 1% chance at 300, still linear increase in difficulty (no curve).

Pre-nerf, the average total number of combines to get from 250 to 300 would be (1 / 0.05) * (300 - 250) = 20 * 50 = 1000.

Post-nerf, at any given skill level x, we have a percent chance of skill up per combine of (25 - (2x / 25))%.

The average total number of combines to get from 250 to 300 is now SUM[x=250,299]{1 / ((25 - (2x / 25)) / 100)} = 2072, a 107.2% increase.

My point is this: A 20% average increase in the number of combines per skill-up is not the same as a 20% average decrease in the chance of skill-up per combine when there is a declining slope. The latter is much, much worse. Many of the posts I've seen are indicating that SoE has made the latter change instead of the former.

[Phantron]

If you've a linear scale from 4% at 250 to 1% at 300, that is 2.5% chance to skill up on average, which is NOT 20% harder than before. It's 100% harder.

You're confusing with the fact that 100% less chance of skilling up (which is to say, never) is obviously not the same as 100% more combines needed to skill up. Even in that case, if the chance to skill up decrease by 20%, that'd be from 5% to 4%, which means 25 combines would be needed on average for each point, or 25% more combines are needed. You're basically saying that the chance of skilling up is now 2.5% with nothing to back it up.

All the patch said was 250-300 is 20% harder. It is assumed this means 20% more combines on average, and it actually makes no difference whether the skillups are uniformly distributed or not. Before the patch you'd expect 1000 combines to get 50 points. Now you'd expect 1000 + 20% = 1200. Even if you interpret 20% harder as 'chance to skill up is 20% lower', that still gets you 1250 combines.

Further the assumption that skill up scales from 7% at 200 to 1% at 300 is wrong because it is said that 200-250 is unchanged. 7% at 200 to 4% at 250 yields a 5.5% average skillup for this interval, so 200-250 would actually be easier, but it's not.

[Twistagain]

Phantron,

I think the point is that we know what we SHOULD expect from "a 20% increase"...but is that accurate wording in the first place?

It is very believable that someone at SOE would decrease the skillup chance by 20% and someone else at SOE would translate that as "20% harder" and put it in some documentation.

For instance, if we assume that there was a 5% chance to skillup before, and we make it 20% less, we now have a 4% chance. But that means that it now takes (on average) 25 attempts, where it used to be 20. That's a 25% increase in skillup attempts. So, is it 20% harder? or 25%? And what does the person writing the documentation think?

It may all just be a big misunderstanding between what the code DOES and what was communicated.

[Phantron]

A 20% decrease in chance of skilling up translates to 25% more combines. It is possible this part was misunderstood but that's still 25% more combines as opposed to the commonly assumed 20% figure and does not explain the apparent difficulty in skilling up (which I haven't experienced).

There's really only 2 possiblities here really:

1. Everything's working fine, some people just have bad day.
2. Everything's not working fine, and it is a lot more than 20% harder.

Also the actual distribution of the skillups do not matter, because unless you think the distribution targets you specifically, you're still going to get the average number of skill ups needed, on average. It is possible that a nonuniform distribution makes the bad luck even worse but it makes the good luck better too. For example we can assumed a totally whacked scale where 250-299 = 100% success, 299-300 = 1/1150 chance of success. This gives 1200 combines on average to skill up. There is a 1% chance that someone on the last point will fail 5000 times in a row (or more) before skilling up, while the chance of using 5000 combines for a normal 1200 is astronomically small under a uniform distribution (approximating it as 100 per point for 50 times, it's 3X10^-93). So yes the bad breaks can be worse, but so can the good ones. Naively 1% of the people will skill up the last point in under say 20 combines, which means 70 combines for 50 points, which again would have some impossibly low chance under a uniform scheme but is possible with this distribution. In the end, it still all balances out since you're as likely to be lucky as unlucky, unless you know you're going to be unlucky somehow.

[Zolina]

Phantron, the core problem is that we are having lower odds on far more combines under the new system. The average odds of a skill-up at a randomly chosen skill-level from 250 to 300 are only 20% worse, but the average odds of a skill-up at a randomly chosen combine are 50% worse, because there are far more combines at the 1% 300 end than at the 5% 250 end. This means it's taking us twice as many combines to get to 300.

I'll list some data points and the average number of combines to skill up at each point:
skill 250 - 20 combines
skill 260 - 24 combines
skill 270 - 29 combines
skill 280 - 38 combines
skill 290 - 56 combines
skill 299 - 93 combines
We breeze through the easier combines in no time but get stuck at the harder combines for a much longer time. 93 combines at the 1% skill-up percent and only 20 combines at the 5% skill-up percent. That's why there's a 100% increase in the total number of combines in the new system.

[Melenden]

There are two possible expectations of the run from 250-300:

1. 20% more difficult means 20% more combines.

This has already been hashed out. If we assume a flat 5% rate then we would expect 1000 combines on average to get to 300 and 1200 post-patch.

2. 20% more difficult means that the average skillup chance across the entire range is decreased by 20%.

For simplicity's sake let's take a linear sliding scale from 5% at 249 to 3% at 299; across the entire scale there is an average 4% chance to get a skillup. In this case, each skillup decreases the chance of getting the NEXT skillup by 2/50 or .04%. Thus the skillup chance at 250 is 4.96%, at 251 is 4.92%, etc.

This means that the total number of combines necessary (assuming failures are negligible) is equal to the sum of all 1 / (0.05 - 0.0004x) where x increases in an integer fashion from 0 to 50. The sum of this series is 1304.

Either way you interpret the remark, there should be an approximate increase of 200-300 combines TOTAL to get from 250 to 300 (with the assumption that the base rate is 5%). Even if the base, in reality, was slightly lower (4.5-4.7), there should still be no more than 400 additional combines to the entire series.

edit to add: on a per skillup basis, we're looking at a theoretical shift from 20 combines per skillup on average for the entire range, to 26 (with a minimum of 20 at 249 and a maximum of 33 at 299).

The question is, is this what we are seeing?