Difference between revisions of "Source 1 (Source)"

Latest revision as of 15:14, 14 June 2009

As I'm sure you all know, the repair system has been very different in the past months, since at least the time of Publish 7. For those of you who don't, here's the deal. Before the change, master weaponsmiths (and I'm presuming armorsmiths and such too) had almost a perfect repair rate. That is to say, on nearly every attempt we'd get "minor blemish" repairs, which only slightly reduced max condition, and only very rarely would the weapon fall apart, and become useless. Since the change, the weapon is falling apart more often than the "minor blemishes" repair, but the "max condition reduced" repair seems to be severely dominating, statistically.

This change was never in the patch notes. The change has never been acknowledged by the dev team. In fact, at the beginning of April, the WS correspondent relayed to us the following message:

Nothing changed with the success rates for repair kits.

A 25 sample set, split between different tool qualities and tailor skill levels is way, way too small a data set for this type of statistical analysis. The randomization chances throw too much noise into the test. You would need at least 100 samples per test case to have a reasonable chance of showing statistical trends. When analyzing these types of situations I prefer tests with several thousand samples myself. Thanks for the info though.

My personal feelings about the tone of this response aside, dismissing any possibility of a problem is premature, and somewhat insulting. Noting that the particular data provided was insufficient is understandable, but when a majority of the community feels there is something severely wrong with the system, I don’t think it’s too much to ask to have it looked at by those with access to far more information than we.

Regardless, I took it upon myself to run some data on weapon repair under controlled circumstances. So, here’s the report.

Assumptions: According to dev reports, fellow WS reports from before the change, and conventional wisdom, the following should influence repair success rate:

• Weapon Condition • Weapon Complexity • Profession (bonus based on being a WS) • Repair Tool Quality • (?) Repair Skill Tapes (to my knowledge, there has been no evidence that these help at all, but since I have none, I was unable to use them in the test anyway) • (?) Standing before a crafting station (I highly doubt this one, but it was there, so I figured it couldn’t hurt, so all repairs were made in front of a 40.67034 quality WDG crafting station)

In an attempt to show that even in a “best case” scenario, or very close to it, repair rates are still far less than satisfactory, I attempted to maximize these points where I was able. In that spirit, 90.69399% repair tools were used. I chose not to use 99.xx % repair tools because the difference from 90-99 shouldn’t be immense enough to illegitimate the data, and I was concerned about what I would call “fringe effects,” in which the very highest and lowest numbers can sometimes expose flaws in the programming due to the fact that such numbers may not have been taken specifically into account during the coding process.

Repair

When repairing an item, the outcome may be one of 4 possible repair types. In order from best to worst:

• The item is repaired with only minor blemishes (MB, 5% max condition decay) • The item is repaired, but the max condition is reduced (RBR, 20% max condition decay) • The item is only marginally repaired, and the max condition is reduced (MR, 35% max condition decay) • The item falls apart in your hands (F, condition goes to 1/1)

1st run

The first group of repairs consisted of a factory run of CDEF Carbines that were all used for approximately 5-50 points of damage. The weapons were all repaired, then brought out again, abused, then repaired again, up to 4 times on some of them. Repair rates showed no fluctuation as weapons were continually repaired.

Of 249 repairs:

MB: 57 (22.9%) RBR: 180 (72.3%) MR: 10 (4.0%) F: 2 (0.8%)

These were compared to the repairs made by another player, who has no WS skills. He did however, have Engineering IV, Domestic Arts I, and Novice Armorsmith.

Of 50 repairs:

MB: 10 (20%) RBR: 34 (68%) MR: 5 (10%) F: 1 (2%)

The smaller, comparative run by the non-WS was done with only 50 weapons, and so the specific percentages are not to be taken as absolute. They do suggest that repairs by non-WS’s tend away from RBR, and toward the lower two, but with such a small rate even on the larger sample set (only 10 MR’s, and 2 F’s out of 250), it is difficult to substantiate this claim. However, one thing is absolutely certain: if there is any difference at all between the repair rates of a WS and a non-WS, it is nominal at best, and seems to have a larger effect on preventing failures than improving the average repair rate.

2nd run

The other group I tested repair on was a random selection of looted weapons. Because they were looted, these weapons had randomly distributed conditions out of a maximum out of 750. The idea of this run was to test a variety of different weapon types (complexities), with a variety of initial condition ratings and to compare these numbers against those of the controlled system (the CDEF test) in which neither complexity nor condition was a factor in determining repair rates.

Of 139 repairs: 1st run comparisons (249):

MB: 26 (18.7%) (22.9%) RBR: 91 (65.5%) (72.3%) MR: 16 (11.5%) (4.0%) F: 6 (4.3%) (0.8%)

Clearly from these numbers, the looted weapons have a lower repair rate than the “best case” CDEFs. As pointed out above, though, the differences between these rates may be from either item complexity, item condition, or both. I first analyzed the effects of item condition, however, by sorting out the repairs done on weapons that had started with less than 50% condition (the first point at which the game system sends the user a message about the item’s decay).

Of 84 repairs (condition under 50%):

MB: 16 (19.0%) RBR: 53 (63.1%) MR: 13 (15.5%) F: 2 (2.4%)

These numbers are surprisingly close to those of all looted weapon repairs, suggesting that item condition actually bears no influence on the outcome of the repairs, especially since these specific repairs showed no systematic difference as compared to the whole of looted weapon repairs (MB almost identical, RBR below, MR far above, F below).

I then looked at the effects of item complexity by selecting out all the CDEF pistols, carbines, and rifles (which should all have the same complexity) and again comparing these numbers against those of all looted weapon repairs and of the 1st run CDEFs.

Of 31 repairs (looted CDEFs):

MB: 6 (19.4%) RBR: 23 (74.2%) MR: 2 (6.5%) F: 0 (0%)

Granted, a set of only 31 repairs isn’t really sufficient for an definitive conclusions, it is worth noting that these numbers are considerably closer to those of the 1st run CDEFs, than to the looted weapons. In other words, the looted CDEFs (with random condition), had approximately the same repair rate as those with near perfect condition. This suggests again that item condition has no real effect, and that the difference between the repair rates of the 1st run and the 2nd was due to item complexity.

Summary

• Weapon Condition – It is unlikely that condition plays any part in weapon repair rates. If it does, the effects are subtle, and nearly unnoticeable. • Weapon Complexity – It seems that higher complexity items show worse repair rates, assuming of course that looted weapons have complexities (which is by no means certain, but if they don’t, looted weapons would show a lower repair rate for an unknown reason). • Profession (bonus based on being a WS) – The repair rates of a Master Weaponsmith, seem to show no significant bonus as compared to those provided by someone with Engineering IV (as well as Novice Armorsmith, but that should make no difference to weapon repair).

The data shows that many of the factors that were believed to affect item repair make NO difference at all, despite developer claims to the contrary. Everyone who was around before publish 7 knows how repair used to be, and this clearly is not the system now in effect. Even when certain conditions showed variations in repair rates, the differences were nominal at best, and unlikely to even be noticeable without recording the data over time.

In my opinion, we should see huge improvements in a “best case” scenario, and dramatically worse in a “worst case.” I don’t think it would be unreasonable to give upwards of a 65% MB rate for a master WS, with good repair tools, working on simple weapons, with low degradation. On the other hand, a soldier attempting to make a repair on particularly complex weapons, run into the ground, with an average or worse tool, should have a majority MR rate, with perhaps even as much as 20% failure rate. These are only my opinions, of course, and the specific numbers would need to be debated and the consequences weighed. However, I feel that a more fleshed out repair system (taking several important factors into consideration), would make repair more engaging, raise the importance of durability experimentation, and increase the turnover of “uber loot” which leads to more balanced play in both PvE and PvP.

And so ends my report. There is more than enough evidence here to show that something is dramatically different with repair rates, as compared with what all of us remember. “Nothing changed with the success rates for repair kits” isn’t going to cut it anymore, and we deserve to know what’s going on.

• • • Special thanks to my guildies and others on Intrepid who helped me with these tests. I couldn’t have done it without you.