As stated in the 'Cat DPS guide for dummies', at around 400-500 ArP it becomes a better stat than Agi. I currently have 314 ArP, but with regemming I could reach ~550. I presume that regemming should increase DPS, but I want to make sure before I invest a small fortune.

So, should we regem to ArP as soon as we can get to a total ArP of 4~500 (including gems), or should we start using ArP as soon as we have 4~500 ArP without gems, with the gems adding to this?

I have run myself through simcraft and regemming to ArP seems to lower my DPS, but perhaps my method is faulted. I used simcraft to pull myself from the armory (./simc armory=eu,turalyon,Zirahe save=zirahe.simc) and used the 'gear summary' entries to model gear. To test how I would look after regemming, I reduced gear_agility by 224 and increased gear_armor_penetration_rating by 224. Using both optimal_raid=0 and optimal_raid=1, my dps after regemming was about 10-20 lower.

Comments about simcraft usage or thoughts about regemming are welcomed.

This is a terrible way to think about the situation. FBs are nearly worthless post-4t8: it's a 100 DPS increase, or only 50 if you don't spec FA. And the more you veer into focusing on FBs, the more likely you are to do it too much and lose Rip uptime. In other words, it doesn't matter if your CP generation falls a little bit, since you'll have plenty of CPs for the "required" finishers.

Just in theory, 66 (.6) % crit raidbuffed is nice for cb generation, because you have the chance of 2 of 3 hits being crits, which results in 5cb. Any additional crit would alter your damage but wouldnt (theoretically)result in more cbs.
You would alter your chance of getting your third cb style critically hit an gain a 6th cb which wouldnt add in for your finisher.

Till you reach this amount of crit (sorry for my english, dont know how to say that) agility is more important than arp, unless you really have a lot of arp or GT/MR.

"In theory," more CP will always mean fewer specials needed for 5 cp. Just because 2/3 of your attacks crit on average doesn't mean that 66.6 is some magical value where CP generation from crits loses all value. The value of crit does, of course, diminish as the likeliehood of reaching 5 cp in 3 attacks increases, but shooting for 66.6 on the dot is just silly.

I didnt say at 66% crit looses its value completly, but regarding cp gain it has fewer value the more you have above this theoretical number.
But in the end all we have are numbers to discuss...

First post here, hope it is helpful

Making the assumption that finishers are only used at 5 cps

Also assuming that the only way to waste a combo point is if you crit at 4 cp's

Doing some quick math with a crude calculator

There are only 3 ways to obtain 4 combo points,

crit chance is calculated as a decimal value of the percentage 60% = .6

4 non-crits (1-crit)* (1-crit) *(1-crit)* (1-crit)
+ 1 crit 2 non-crits (1-crit)*(1-crit)*(crit)*3
+ 2 crits (crit*crit)

adding these, then multiplying them by your crit chance will give you your chance to waste a combo point

(total*crit)
crit%-------Chance to waste combo point
30%------- 23.1%
40%------- 28.8%
50%------- 34.3%
60%------- 40.4%
70%------- 48.0%
80%------- 59.0%

So you can see that the chance to waste a combo point rises at an exponential value, and while 66% seems like a good middle ground, there's actually no specific value at which crit "starts to diminish".

it would be interesting to see some calculations on the dps value of a combo point, and the ratio of combo points gained to combo points wasted.

upon further evaluation your chance to gain a combo point goes up at a near exponential rate as well, in reality, there probably isn't a "starts to diminish" point on crit. From early readings, it seems like the opposite is true, crit actually goes up in value (not counting FB soft cap);

Edit: Note that the "Chance to waste combo point" is per finisher

Please could you explain how you calculate the chances of wasted cps, i didnt get it, thanks!

A few comments, specifically addressing these points but also the post as a whole:

- While many strategies tell you to "aim for 5 CP for optimal CP usage, you might instead aim for 4 CPs on SRs and never "waste" the CP due to crit. Or, aim to achieve equal uptimes on SR/Rip, which is in the 3-4 CP range. These strategies minimize combo point waste in the manner you describe. It's still true for Rip, though.

- Less interesting than maximizing CP usage by reducing waste is minimizing the average number of attacks to reach the 5 CP goal. The minimum, of course, is 3, and the maximum is 5. ~66.7% is the red-herring number where you'd naively expect to see two crits in three attacks, and thus average pretty close to three CP-generators per cycle. This isn't the case, of course. The goal is to minimize non-finisher attacks.

- The value of a CP is really hard to define. It might be 1/5th of the difference between an FB and a shred, if you FB a little, and waste no CP. Although to truly not have FB impact your Rip uptime, you need lots of spare CP, and you're likely to waste some. In other words: any exact value you place on CP will likely be lost in the complexities of managing your DPS in-game.

I'll bring more math later, and I might even be able to track down a pretty attacks-to-5-CP-versus-crit graph.

I support this fully. Actually it's even more pronounced. Not counting the potentially wasted CP due to getting a crit with 4 CPs, Savage Roar becomes also less efficient (in terms of Duration / CPs spent) the more CPs you put into it:

1CP -> 14Sec -> 14 Sec/CP
2CP -> 19Sec -> 9.5 Sec/CP
3CP -> 24Sec -> 8 Sec/CP
4CP -> 29Sec -> 7.25 Sec/CP
5CP -> 34Sec -> 6.8 Sec/CP

This is partially mitigated by the energy cost of SR itself, we can see if it makes sense if we can attribute an energy cost to a CP.
I took an old combat log, checked how many mangles, rakes and shreds I cast, calculated the total energy spent / attacks done, adjusted for 64% crit rate, ended up with 23.72 energy / CP
(WARNING: single combat log, totally unscientifical, your result may vary wildly, etc. Also, I didn't model the manglebot case and the fact that OoC procs are mostly saved for shreds, reducing its average cost, further complicated by the fact that we sometimes have to shred at 5CPs. It probably brings the cost down to about 21.5 per CP)

So how many seconds of Savage Roar buff does 1 energy buy? By adding up the average CP energy cost and SR's own 25 energy cost, we have:

1CP = 25+23.72 = 48.72 = 0.287

2CP = 25+47.44 = 72.44 = 0.262

3CP = 25+71.15 = 96.15 = 0.250

4CP = 25+94.87 = 119.87 = 0.242

5CP = 25+118.59 = 143.59 = 0.237

SR is thus more energy efficient when used with a low amount of CPs, 1 (and 2 if you crit, ofc) being the best (21% more effective than 5CP).
I tried changing the ability mix, crit rate, adding 1 second of SR overlap, and it's always most effective with the lowest CPs.

Small edit:
Due to crits being discrete (you either crit or don't) I made 2 additional tables for the cases of getting 1 or 2 crits while building those CPs:

1Crit (as early as possible)
1CP is ofc not possible
2CP 0.390 sec/energy (1 crit)
3CP 0.331 sec/energy (1 crit, 1 normal)
4CP 0.302 sec/energy (1 crit, 2 normal)
5CP 0.284 sec/energy (1 crit, 3 normal)

2 Crits (as early as possible)
1CP is not possible
2CP 0.390 sec/energy (only 1 action was used, so only 1 crit really happened here)
3CP is not possible
4CP 0.400 sec/energy (2 crits)
5CP 0.354 sec/energy (2 crits, 1 normal)

So, the case of 4 CP obtained with 2 crits in a row is the most energy efficient (no shit Sherlock), but it's _barely_ better than 1 single initial crit (0.4 vs 0.39). Thus, if your first action crits, you should stop right there and SR with those 2 CPs. If your next action crits as well, grats, you'll use those on the next finisher instead of such a marginal improvement in SR effectiveness.
What happens if your first action doesn't crit becomes too dependent to your own crit chance and ability mix, so you'd rather use other, non-energy related deciding factors (as in ensuring the SR won't expire too close to a rip, or evaluating if you could squeeze a FB or not, or if there will time with no boss contact, if you can pool energy or not, etc)

Thanks, obviously i was wrong

Not sure if this is what you're saying, but only using 1 attack to cause a savage roar also seems like it would interrupt your cp building more often, unless you can get into some rotation like

1/2 point SR
5 point rip
1/2 point SR
5 point FB

it seems that there are some disadvantages to not having the long SR uptime, although, i could be completely wrong on this, i'll come back after some testing