[Mini-Guide] Mabinogi Repair % System Is a Lie?

[Jahrakal]

%The backbone of this guide was originally posted on the Nexon Official Forum%

%English is not my first language, I apologize in advance for grammar/spelling errors%



How often have you failed your repairs and said "OMG MABI REPAIR RATES ARE BULL****!"

This is a mini-guide for people who are curious about the repairing mechanics in the world of Mabinogi.

Repair Success Rates - Oblivious Facts Behind Repairing Mechanics

Even at 98% success rate, you still fail sometimes a few times in a row. This is not % system's fault, but your lack of knowledge on how probability works.

Here's a simple example to aid you with basic concept of how repair rates work. In this example, I'll discuss 95% repair rate.


Think of a box full of 100 golf balls. 5 of these golf balls are defective (let's call these red balls).


What are the odds of getting a red ball when you randomly pull one of the golf balls out from the box? 5/100.
What are the odds of getting a golf ball? 95/100, hence, the repair rate.


Even with 98%, eventually (with replacements), you will pick a red ball at least once in the long run. Sometimes you get un/lucky (depends on how you want to look at it), you get another red ball after your first pick. Note that 5/100 is only 1/20, so out of 20 tries, on average, you will fail once.

It feels like you are losing out a lot simply because the number 95% and 98% look promising to the eyes. I am willing to bet that you do not remember exactly how many times you succeeded repairing since you started playing, but you roughly do remember how many times you have failed the repairs.

If one was to write out a chart of success/failure from one specific NPC since he first started playing, the average will be close to the repair rate of that NPC.


Difference Between Perfect Repairs and Point-by-Point Repairs.

If you do perfect repair (meaning attempting to repair all points at once), your success rate drops dramatically, resulting in almost a guaranteed failure. This is why a lot of players use point-by-point repairs.

Example:

Let's say you're trying to repair 0/10 dagger, blessed/non-blessed is irrelevant for this discussion.

If you do point-by-point, your repair sequences and rates are like this.

98% -> 98% -> 98% ->.... -> 98% -> 98%


If you choose perfect repair, it looks more like this.


(98% 98% 98% ... 98% 98%)



On perfect repairs, the system overlaps the subsequent repair rate followed by the first attempt.


For easier understanding:

[98%] -> [98% of 98%] -> [98% of 98% of 98%] -> .... -> [98% of 98% of 98% of... 98%]

Each arrow representing each point in durability when you're using perfect repair.

See the problem here? The chances decrease the more durability loss you have. Therefore, items with 0/5 durability have way higher success rate than the ones with 0/30 durability, using the perfect repair.

Example: At 98% repair, using perfect repairs

Dura : Success Rate
0/1 : 98% chance to fully repair (This is the same as point-by-point repairs)

0/5 : 90% chance to fully repair
0/10 : 81% chance to fully repair
0/15 : 74% chance to fully repair
0/20 : 67% chance to fully repair
0/25 : 60% chance to fully repair
0/30 : 55% chance to fully repair


Following is an analogy irrelevant to probability, but explanation of the concept of why the rates are like that:

Think of 30 cars trying to park in 30 different spots all lined up side by side all at once. - Perfect Repair
Think of 30 cars lining up in 1 line to fill up 30 spots 1-by-1. - Point by Point


To sum it up, if you attempt perfect repairs at low durability, you are pretty much guaranteed to fail a point.

I will not mention the effects of Holy Water and rainy days, as they have very little to do with my topic.

Cheers.

[EndlessDreams]

What does low durability have to do with anything?

You should just say that the more durability you repair, the more likely you going to break a point.

It is kind of funny that you have to show all that probability just to show that 98% repair isn't guarantee repair.

[Sam]

Hopefully this isnt true, although it would make sense, and why even Edern always seems to fail at least a point of durability! D:

[Jahrakal]

What does low durability have to do with anything?

I see that you weren't able to comprehend the low durability part. Perfect repairs with low durability items have lower success rate than the actual success rate.


You should just say that the more durability you repair, the more likely you going to break a point.

Again, read it over carefully, and maybe you'll see that my mini-guide simply goes into more details explaining the mechanics.


It is kind of funny that you have to show all that probability just to show that 98% repair isn't guarantee repair.


Title: Oblivious Facts Behind Repairing Mechanics

You might as well say all the guides in the forum are for nothing.


I'll say this again. This is the mechanic behind repair success rate. I never said anything about 98% being a 100% guaranteed repair. It seems as if you got nothing better to say than to whine aboout the details I've provided. I find that funny.

[Shironi]

Ferghus.

1. 90%
2. 81%
3. 72.9%
4. 65.6%
6. 59%
7. 53%
8. 47.79%
9. 43%
10. 38.7%

Moral of the story, don't repair at ferghus.

[Cide]

Hopefully this isnt true, although it would make sense, and why even Edern always seems to fail at least a point of durability! D:

its basic math. of course its true.

[Chiri]

Some other things you can cover:

The misconception that...

... if you fail once, the chances are higher that next time that you will succeed
... if you fail 3 times in a row, then the stated probability must be false

[Andy-Buddy]

Well, I got bored.

Repair.xls (http://www.mediafire.com/?xize3atogtj)
^That is an excel spreadsheet where you can just put in your own values and repair %s to check what your chances are.
Blessing is in there too.


I might never do the % of losing a certain amount of points, depends how bored I get.

Rep would be appreciated.:lol: :3

[Yogurticecream]

I swear by point-by-point repairs.
And I also do not do repairs in one row, I'll usually run to get my items repaired the moment they go down by one or two durability.

I do lose durability sometimes, but that's to be expected.

[EndlessDreams]

I see that you weren't able to comprehend the low durability part. Perfect repairs with low durability items have lower success rate than the actual success rate.



I see you aren't comprehending the bigger picture.

If you have 0/8 weapon, you repair the whole thing in one visit at NPC, you have 8 attempts of repair.

If that same weapon, you let it down 1 point to 7/8, and repair it. Then, you repeat it 7 more times over the course of your dungeoning, you will repair the same amount of points.

The probability is the same.

You missed my post completely. For someone who wrote all that, you didn't even look at the bigger picture.

[asuran]

Put this simply.

You want to know what's the chance to repair your weapon at 98%? do this

0.98^n

where n is the number of durability points you want to repair.

Just to break down some % rates.

Chance on getting 50 durability points repaired in a row.

98% -> 0.36416968008711706521737398145318 = 36.4%
95% -> 0.07694497527671332927429437909584 = 7.6%
90% -> 0.00515377520732011331036461129766 = 0.05%

Even a increase of just 1% on top of 98% would sky rocket the chance to get 50 repairs in a row to 60%. I don't know how much % chance does holy water add but if it does add 1%, it has a huge impact on a 98% repair.

And what you said about full repair and repair one point at a time makes no sense, people repair like that because if they FAIL they can re-bless the item and thus don't lose the extra % chance you get when repairing a blessed item.

If anyone doesn't understand this it's because either you are too young and haven't had probabilty and stadistics class or you had it but you failed it.

More info here - Simple random sample - Wikipedia, the free encyclopedia (http://en.wikipedia.org/wiki/Simple_random_sample)

[Raj]

Ferghus.

1. 90%
2. 81%
3. 72.9%
4. 65.6%
6. 59%
7. 53%
8. 47.79%
9. 43%
10. 38.7%

Moral of the story, don't repair at ferghus.

Totally.

Sorry, I found most of this common sense.
It would've been pro to say-

'I know Ferghus is cheap, but don't go to him.
Enough said.'
xD

[Jahrakal]

I see you aren't comprehending the bigger picture.

If you have 0/8 weapon, you repair the whole thing in one visit at NPC, you have 8 attempts of repair.

If that same weapon, you let it down 1 point to 7/8, and repair it. Then, you repeat it 7 more times over the course of your dungeoning, you will repair the same amount of points.

The probability is the same.

You missed my post completely. For someone who wrote all that, you didn't even look at the bigger picture.

How is that a bigger picture? Where exactly in my guide did I state that those two methods of repairs are any different? I'm promoting point-by-point repairs and pointing out the misconception of perfect repairs.

Thank you for your input nontheless, but the tone of your first post wasn't very welcoming for something I put my own time into.


@Andy-Buddy

How neat =)

[Jahrakal]

And what you said about full repair and repair one point at a time makes no sense, people repair like that because if they FAIL they can re-bless the item and thus don't lose the extra % chance you get when repairing a blessed item.

No, point-by-point repair and perfect(full) repair are different in nature. That is just a side benefit of repairing by point-by-point, but point-by-point does more than that.




If anyone doesn't understand this it's because either you are too young and haven't had probabilty and stadistics class or you had it but you failed it.

More info here - Simple random sample - Wikipedia, the free encyclopedia (http://en.wikipedia.org/wiki/Simple_random_sample)

Hi, Math Major here with 6 distinctive courses taken in Statistics and Probability.

Of course this is a common sense, but you will be surprised at just how many people use perfect repairs because they think the mechanics behind the point-by-point repairs and that of the perfect repairs are the same.


- - - - -

Edit: I apologize for the tone of my posts being defensive. Regardless I'm open for your feedbacks. Keep them coming.

[Andy-Buddy]

Holy Water cuts the % of loss in half.
i. e. 90% -> 95%
95% -> 97.5%
98% -> 99%

We were just looking at the chance of a perfect repair, at different points of time. Blessing failures, and losses are not counted in what we are looking for.

[Cide]

Holy Water cuts the % of loss in half.
i. e. 90% -> 95%
95% -> 97.5%
98% -> 99%

We were just looking at the chance of a perfect repair, at different points of time. Blessing failures, and losses are not counted in what we are looking for.

is there any source out there to actually confirm this? i've yet to see a solid source but i've heard this a lot.

[Syrphid]

Difference Between Perfect Repairs and Point-by-Point Repairs.

If you do perfect repair (meaning attempting to repair all points at once), your success rate drops dramatically, resulting in almost a guaranteed failure. This is why a lot of players use point-by-point repairs.

Example:

Let's say you're trying to repair 0/10 dagger, blessed/non-blessed is irrelevant for this discussion.

If you do point-by-point, your repair sequences and rates are like this.

98% -> 98% -> 98% ->.... -> 98% -> 98%


If you choose perfect repair, it looks more like this.


(98% 98% 98% ... 98% 98%)



On perfect repairs, the system overlaps the subsequent repair rate followed by the first attempt.


For easier understanding:

[98%] -> [98% of 98%] -> [98% of 98% of 98%] -> .... -> [98% of 98% of 98% of... 98%]

Each arrow representing each point in durability when you're using perfect repair.

See the problem here? The chances decrease the more durability loss you have. Therefore, items with 0/5 durability have way higher success rate than the ones with 0/30 durability, using the perfect repair.

Example: At 98% repair, using perfect repairs

Dura : Success Rate
0/1 : 98% chance to fully repair (This is the same as point-by-point repairs)

0/5 : 90% chance to fully repair
0/10 : 81% chance to fully repair
0/15 : 74% chance to fully repair
0/20 : 67% chance to fully repair
0/25 : 60% chance to fully repair
0/30 : 55% chance to fully repair


Following is an analogy irrelevant to probability, but explanation of the concept of why the rates are like that:

Think of 30 cars trying to park in 30 different spots all lined up side by side all at once. - Perfect Repair
Think of 30 cars lining up in 1 line to fill up 30 spots 1-by-1. - Point by Point


To sum it up, if you attempt perfect repairs at low durability, you are pretty much guaranteed to fail a point.

I will not mention the effects of Holy Water and rainy days, as they have very little to do with my topic.

Cheers.

First, thanks for posting your guide. Always good to see new stuff.

I don't understand the quoted though. Are you saying that full repair has lower dur loss rate? I don't think that's true. Full repair is just single point repair lumped together. On average you would still lose the same amount of dur.

All you're doing is showing how 0.98^n drops down over time. n=30 for both 30 pts full repair and 30 pts point repair.

[Phunkie]

Full repair and point-by-point repairs are the same.

The reason why people use point-by-point repairs is that they can re-bless the item if the repair fails.

You cannot re-bless the item when you fail in a full repair. So if you're repair 4 durability points and you fail on the second (using full repair), you have a lower chance of repairing the next points after that.

Full repair example:

Item is blessed
-> (Repair 1 pt.) Blessed
-> (Failed 1 pt.) Unblessed
(Lower chance of success now since item is unblessed)
-> (Lower chance of repair) Unblessed
-> (Lower chance of repair) Unblessed


However, in a point-by-point repair, you can stop in-between repairing points to bless, just in case you fail 1 pt.

Point-by-point example:

Item is blessed
-> (Repair 1 pt.) Blessed
-> (Failed 1 pt.) Unblessed
(Bless the item, proceed on repairing)
-> (Higher chance of repair due to bless, but you can still fail again)
-> (Bless again if you fail, continue on if you don't)
-> etc., etc.


With that said, if you never bless your items, point-by-point repairs and full repairs are completely identical. It doesn't matter which one you do.

0.98^n applies to any of them in that case because the probability is the same.


The OP should include a section on blessings and Holy Waters.

[Raj]

Did I see a double post?!~

And - Erm, I don't like all the math.
>.>

[Phunkie]

It's pretty much just punching numbers in your calculator. That's not real math.

[Raj]

Oh. o.o
I don't have a calculator.

[Rime]

Full repair and point-by-point repairs are the same.
...
...
...
0.98^n applies to any of them in that case because the probability is the same.


The OP should include a section on blessings and Holy Waters.

This is only true if you're looking at the situation from before any points are repaired.

If you are repairing your weapons point by point your frame of reference shifts to before the repair of the next point, but after the repair of the last point, so the success rate of the next 1 point of repair is always the success rate mentioned by the black smith.

For instance, let's say I'm repairing a 0/3 broadsword at Erden, un-blessed. The chance of success for the first point is 98%. I click "repair one point" and it succeeds. My frame of reference is now after the successful attempt of repairing 1 point, so the success rate of the next 1 point from my current frame of reference is also 98%. Let's say the second point also succeeds. My frame of reference is now between the second and third attempt in which the first and second attempt succeeded. It follows now that the success rate of the third point from my current frame of reference is still 98%.

Full repairs are, by nature, different because the game calculates the probability of success of all points at once.

Let's take the same situation from above and assume that I'm much too lazy to click three times, and have resigned to only do so once for a full repair. Below, I'll show you how the system will treat the well-being of my broadsword.

Scenario 1: Loss of no points There's only one way this can happen, and that is if no points fail.
.98 x .98 x .98 = .941192

Scenario 2: Loss of one point. There are three ways in which one point can be lost. The first point can fail or the second or the third. For the sake of simplicity, I'll write it out this way:

[Fail][Success][Success]

[Success][Fail][Success]

[Success][Success][Fail]

The probability of each one of these cases occurring is:

.02 x .98 x .98= .019208
.98 x .02 x .98= .019208
.98 x .98 x .02= .019208

The probability of losing one point taking these three scenarios into consideration is, simply, the sum of the three which equals .057624

Scenario 3: Loss of two points. There are three ways in which I can lose two points. I'll write them out below:

[Fail][Fail][Success]

[Success][Fail][Fail]

[Fail][Success][Fail]

.02 x .02 x .98 = .000392
.98 x .02 x .02 = .000392
.02 x .98 x .02 = .000392

Chance of losing 2 points = 3x .000392 = .001176

Scenario 4: Loss of all 3 points There's only one way this can happen. It is, obviously, if none of the points succeed.

[fail][fail][fail]

.02 x .02 x .02 = .000008

Data check:
.941192 + .057624 + .001176 + .000008 = 1
No scenarios were overlooked.

The big picture:
The system calculates all the above scenarios faster than you can muster them, and treats your weapon accordingly. So, to predict the results of the full repair of my broadsword, let's restate the information above as something more understandable.

As calculated from above:
Possibility of no points lost = .941192, which is also equal to 94.1192%

Possibility of 1 point lost = .057624 or 5.7624%

Possibility of 2 points lost = .001176 or .1176%

Possibility of all points lost = .000008 or .0008 %

If I choose "full repair" for my broadsword, there will be a 94.1192% chance that no points will fail and the sword will be at 3/3 dura; a 5.7624% chance that one point will fail, leaving the sword at a paltry 2/2 dura; a .1176% chance that two will be lost, making the sword almost unusable; and finally, a .0008 % chance that three points will be lost, leaving me with a completely broken 0/0 broadsword.

Compare this to point by point repair at Erden, in which the next point will always succeed at a promised 98%.

Take this however you will, cause I'm too tired from typing all this data to give you my opinion. Also, side note: I haven't taken stats in like, forever, so the terminology might be a bit awkward, or even incorrect.

[Taran]

This is not % system's fault, but your lack of knowledge on how probability works.




*yawns at thread*

So explain why i failed 3/4 times (fail-fail-success-fail in a row) at 99% making silks? Either Mabinogi % system is broken or during that minute all odds were against me.

[Mentosftw]

^ Probability. 99% does not equal 100%. That slim chance is still a prominant factor in getting a failure. You have to realize that the difference and distance between 99% and 100% is extremely great.

[Kazuni]

What Phunkie said...

Point by point is for blessing in between in case of failure. There's no % difference in the repair itself.

eg.

98^n isn't a formula for the % of success you get for an all point repair or whatever, it's for both (1 point and perfect), over time. I don't see the point of all these calculations. It's the same.

[Mentosftw]

It's different because of the way the game system calculates it.
Point by point works with odds stacked in the real word where the odds ARE stacked but counted for in real life.
What I mean is that the game system takes each point by the specified repair rate one at a time but the odds of all your points repaired will be dependant (although each point will not affect each other's rate but when all the rates combined will be a single rate).
Perfect is different because the system already calculates the combined probability of all the repairs internally and then applies it to the repair.
Therefore, technically both probabiliy outcomes for point by point and perfect are the same, because they are applied differently, it's different. If you were somehow able to go back in time and replay the same repair (item must be over 1 point of damage), only to either select perfect or point by point, the result would be different.
Sorry for the crappy explaination.

[Kazuni]

No, but still. They calculate it with the correct math, which would initially be the same as if you did point by point.

You may get different results should you go back in time, but if you were to pick a coloured marble out of a bag you might also get different results.

In the end, unless devcat is purposely screwing, which would be hard to prove at any rate, the % should be the same for both point by point and perfect no matter how you calculate it.

[Rime]

No, but still. They calculate it with the correct math, which would initially be the same as if you did point by point.

You may get different results should you go back in time, but if you were to pick a coloured marble out of a bag you might also get different results.

In the end, unless devcat is purposely screwing, which would be hard to prove at any rate, the % should be the same for both point by point and perfect no matter how you calculate it.

Depends on whether or not they round the percentages or the RNG that deals with the randomness, I think. RNGs are never completely random...

[Mentosftw]

Well, as for the going back in time thing, I have a real life example. This deals with two games on a GBA emulator, Super Robot Wars OG 1 and 2. As all of you may know, there is a % chance for all battle encounters. The difference is that OG 1's % calculator always has a fixed outcome for every encounter even with save states that allow you togo back to the moment where you created the savestate or "go back in time". Every time that savestate is loaded, the outcome is always the same even if the percentage is 50%. However, with OG 2, every time you loaded the savestate and replayed the encounter, it would always be different. There is obviously a fixed outcome and not a completely random chance. The same goes with my first time travel example. In that time frame, that outcome will always be the same just like OG1 otherwise, we'd start talking about alternate time branches (I don't want to go there) like OG2 where I can replay the encounters to perfection leading to a game where I avoided everything and missed nothing. To get to the point, since point by point and perfect are two different options, they'd most likely lead to different results. To complement your coloured marble example, if you continiously watched yourself pick out from the same container in that same timeframe, you would find that you past self's hand will always move exactly the same way to always grab the same marble. It would be like there being TWO containers, both with fixed outcomes because of the consistancy of time but a different outcome over the other because of a different option.

[Kazuni]

So what you're saying is that repair has a fixed outcome and % doesn't matter because it's fixed?

And as a response to.. your response.. about the marble thing, it was just an example for the chance of going back in time, and the angle of the hand and possible influences are not considered factors, as repairing in mabinogi is purely electronic and not dependent on outside factors at all.

[Rime]

So what you're saying is that repair has a fixed outcome and % doesn't matter because it's fixed?

And as a response to.. your response.. about the marble thing, it was just an example for the chance of going back in time, and the angle of the hand and possible influences are not considered factors, as repairing in mabinogi is purely electronic and not dependent on outside factors at all.

"Electronic" probabilities are dependent on complex algorithms. In Mento's example, the difference between OG1 and OG2's RNG algorithms is that OG1's must not have varied with passing time. That said, since we cannot determine the working of Mabi's algorithms and whether or not the algorithms significantly affect the repair rate, this whole unblessed perfect repair vs. point by point repair is pretty inconclusive.

[Mentosftw]

So what you're saying is that repair has a fixed outcome and % doesn't matter because it's fixed?

And as a response to.. your response.. about the marble thing, it was just an example for the chance of going back in time, and the angle of the hand and possible influences are not considered factors, as repairing in mabinogi is purely electronic and not dependent on outside factors at all.

In a sense, it IS fixed. Every experiement is always fixed. You just don't know the outcome yet till you try it :P
Let's say a tossed die ends up a 4, that's one experiment. It'll always be a 4 no matter how many times you can reverse time and watch it roll, it'll always be a 4. The randomness of chance only exists the moment an experiment with odds is about to happen. Basically, when you don't know the outcome and only the odds. The odds only show you the ratios of the outcomes. Once it happens, you can say that outcome for that experiment was a one hundred percent "chance" it would happen.
What I'm trying to get at both logically and mathematically is that if you could choose both point by point and perfect at the same time, you would get a different result logically because they are two different options but also mathematically because point by point and perfect presents the same (you can argue differently because of the RNG of mabi) odds but chance will "choose" a different outcome (unless they both end up with the same outcome by chance of course, and that chance of both of them either having the same or a different outcome is another discussion and one dealing with hypothetical situations that can't happen (going back in time, as of now anyway)).

I digress, the main point is that because of the unknown algorithim of this game's system and because of how it deals with "chance", it changes things. At the same time it brings a bit of irony as all this is almost as if the blacksmiths of mabi "choose" whether or not to fail with your equipables.

[Trigger]

This entire guide assumes from the get-go that Mabinogi actually uses a correct probability model, which is a pretty big assumption given that we have no source code to prove it.

Assuming for the sake of argument that Mabinogi actually does use correct probability, six years of mathematics studies doesn't mean your word is law. Tossing out some credentials lends more weight to your words, but it doesn't mean you're infallible. "Nothing by mere authority" ---Royal Society motto


Point-by-point and full repair are identical from a probability standpoint.



My chance of flipping heads on a coin is 50%.

If I flip twice, my chance of getting heads both times is 50% * 50% = 25%.

The chance for each individual flip is 50%. That never changes.



Assuming no Holy Water is involved:



My chance of repairing a point of durability at Ferghus is 90%.

If I repair twice, my chance of repairing successfully both times is 90% * 90% = 81%.

The chance for each individual repair is 90%. That never changes.



All the arguing that has taken place in this thread is nothing but semantics. The probability of repairing a single point is always the same! That probability stacks on top of itself the more consecutive points you repair. You can't thwart the numbers by considering each consecutive repair a new set of calculations; it's the same either way you look at it.

The advantage to repairing point-by-point is what has already been stated -- repairing one point at a time allows you to rebless your gear should a repair fail.

[alphabetical]

Point by point and perfect repairs DO have different approaches but the BEGINNING and END RESULTS are the same and that's what matters. Who wants to know what happens between those two points in time? All we want to know is how long the weapon will last.

Lets say you have a 0/3 dura weapon. If you choose to repair point by point, obviously you're not going to repair it once, and leave with 1/3 dura right? (or maybe 0/2) No, you're going to repair it until it is either 3/3, 2/2, 1/1 or broken.

Repairing point by point, 3 times in a row:
-Chance 3/3 dura left = 85.7375%
-Chance 2/2 dura left = 13.5375%
-Chance 1/1 dura left = 0.7125%
-Chance 0/0 dura left = 0.1250%

Perfect repair has the exact same chances for the 4 different outcomes above. You can't compare the 95% success rate of a single point repair to a ~85% success rate of a perfect repair of 3 points.

[Rime]

Actually, blessed point by point vs blessed perfect repair is inconclusive as well. We still don't know whether or not the increase in repair rate is applied to all points in "perfect repair" or if the bonus is taken off for the rest of the repair once a failure occurs. Since the blacksmith doesn't stop the repair and return your weapon in the instance of a failure, I'm inclined to think that there must be some kind of benefit to perfect repairs other than saving a few seconds.

[Trigger]

Actually, blessed point by point vs blessed perfect repair is inconclusive as well. We still don't know whether or not the increase in repair rate is applied to all points in "perfect repair" or if the bonus is taken off for the rest of the repair once a failure occurs. Since the blacksmith doesn't stop the repair and return your weapon in the instance of a failure, I'm inclined to think that there must be some kind of benefit to perfect repairs other than saving a few seconds.

JP wiki says blessing applies only to the first point in a perfect repair, or until a point is failed (not sure of the wording since I'm taking this from Google Translate, which is inaccurate at best). Given the reputation of the JP wiki I'd be inclined to believe it.

[foogro]

Couple comments here (maybe I missed some of the discussion above):

1) Independence:
We still don't know whether mabinogi repair rates are independent. For example, is your first repair at 98% really 98%? If you have a success, does it affect your second repair rate? Perhaps on average, you get 98%, but making a mistake on one point is likely to cause you to make a mistake on a second point. For example, for Edern to fail consecutively on 2 repairs (on an unblessed item), if each repair is independent, two consecutive failures would occur with expectation once every 2500 points of repair. I'm not sure if that's the case, but I've had him fail 3-4 points on me in a single repair of maybe 10 points. It is possible that there is some correlation.

Take an extreme example: suppose that your weapon is brand new. Edern will repair at 99% if he hasn't failed yet, so he fails on average once every 100 points. However, once he fails, his repair rate for the next 2 repairs drops to 50%. Then he is back at 99% again regardless of whether those points were successful or unsuccessful. On average, this repair rate is (slightly above) 98%, about 98.04%. However, you are very likely to fail more than 1 points at a time.

(This is, btw, one of the intricacies in Warcraft 3 critical rate computation. The first attack actually has very low critical rate, but after each failure to critical, the next hit has a higher probability to critical. Some people actually attack creeps a few times until they get several hits without critical, and go into battle to hit high probability criticals.)

2) Variance

Yay, more statistics! Anyway, another reason you get pissed at repairs is accumulated variance, not just expectation (which has been the sole focus up to this point). Variance expresses the randomness of the repair process. If you repair 1 point at 98%, you have a variance of (1-0.98)^2*0.98+(0.98^2)*0.02, which is approximately 0.02. Standard deviation, which is the square root of variance, is 0.14 points. What does standard deviation tell you? This is not a precise definition, but std dev gives you an idea of the likely range of failures. So basically, the number of points you fail on average per point of repair is 0.02, but you're likely to fail between 0.02 plus/minus 0.14 points.

Given N points of repairs (assuming independence), std dev grows with sqrt(N), so 100 point repairs gives you a std dev of 0.14 points times 10, which is 1.4. Your expectation is 2 failures, but there's a high probability of failing between 0.6 and 3.4 times. Don't be surprised if Edern fails 3-4 points on your sword. If he fails 6 times though, then consider yourself one of the 3% of people who have met great misfortune.

3) If you want to try to build a model of repair correlations (holy water, consecutive fails, etc) it will be extremely difficult because of the variance. You will need to sample maybe 10000 points of repairs to get a fairly accurate model. T_T

[Trigger]

You will need to sample maybe 10000 points of repairs to get a fairly accurate model. T_T

Hence why we simply say it's straight probability and accept whatever misfortune we happen to have while repairing. That, or make long threads trying to explain something we should all have accepted by this time (I mean, come on, it's been two years people <_<).