Marked for Death vs Anticipation for Combat

I decided I want to go into detail about the choice between these two talents. I understand that anticipation feels pretty mandatory for most rogues, and I felt that way too until late into Siege of Orgrimmar. I have since switched to Marked for Death due to a variety of reasons. I think one compelling (and perfectly valid) reason for choosing anticipation is that it simply frees up a keybind. I do not, however, think that makes it simpler to use than Marked for Death. Marked for Death is pretty simple to use, too. With this post I am going to go into detail on exactly how these two abilities impact your dps and your damage breakdown.

Just this morning I decided to run 7 simulations with simulationcraft-548-8 build. Note that this is the most recent live MoP build, not a WoD beta build. That said neither Anticipation nor Marked for Death has changed in Warlords, so the conclusions on these two talents SHOULD be equally valid for the upcoming expansion. You can, however, put this information into practice now.

When I first considered using MfD some time ago, I initially took into consideration only the combo point aspect of the two talents. That alone presents an interesting choice. Choose anticipation and you can do two things. First thing, is that any extra combo points generated by Revealing Strike are not wasted. Anticipation offers this flexibility to save otherwise-wasted combo points. Secondly with Anticipation, you can pool these extra combo points, allowing you to move up to one finisher from no insight to deep insight. On the other hand, MfD gives you up to 5 free combo points every minute, which you can choose to spend in any insight level of your choosing. Well, without even considering the anticipation pooling, we can try and see from a combo point only perspective which is better. In doing this I built a list of all possible combo point generation permutations, then considered finisher frequency coupled with average combo points wasted per finisher.

So lets go to an analysis of the permutations. We’ll group them by conditions of Ruthlessness and Revealing Strike because the conditions triggering these combo generators may or may not be necessary.

We’ll start out by assuming an initial combo point from ruthlessness, and no need to refresh RvS. The chance of getting 4 combo points in 4 Sinister Strikes is 40.96%, 4 in 3 is 38.4%, and 4 in 2 is 4%. This means the chance of getting an extra combo point in this setup is 16.64%. To be more explicit, the chance of overflow on SSx4 is 10.24%, and on SSx3 is 6.4%. With SSx2 it is not possible to overflow combo points. As a sanity check, lets add up all percentages: 40.96+38.4+4+10.24+6.4 = 100.

In the situation where we have a ruthlessness combo point and have to refresh revealing strike, but revealing strike is not the 5th combo point (because if RvS is 5th combo point, chance of overflow is 0%–if you’re at 5, you finish, if 4, you hit RvS), the chance to get 3 combo points in 3 Sinister Strikes is 51.2%, 3 in 2 is 32%. This means the chance on getting an extra combo point is 16.8% (12.8% to get overflow on 3rd SS, 4% chance on overflow on 2nd SS). Sanity check: 51.2+32+12.8+4 = 100.

Next we’re going to look at no ruthlessness combo point or Revealing Strike. The chance of getting 5 combo points in 5 Sinister Strikes is 32.768%, 5 in 4 is 40.96%, 5 in 3 is 9.6%. Chance for overflow is 16.672% (8.192% at 5th SS, 7.68% at 4th, 0.8% at 3rd SS). Sanity check: 32.768+40.96+9.6+8.192+7.68+0.8 = 100.

The probabilities for needing to refresh Revealing Strike but without the Ruthlessness combo point are exactly the same as the Ruthlessness without revealing strike (replace the Ruthlessness CP with RvS, because, again, no overflow with RvS as 5th combo point).

Any time Shadow Blades is down, the overflow is exactly 1 (otherwise there is no overflow). If you include shadow blades, however, you can overflow up to 2 combo points, since everything except ruthlessness generates 1 additional combo point. Also whenever shadow blades is up, the number of builders you will use is exactly 2 with ruthlessness. Without ruthlessness it will be 2 or 3. We will also consider probabilities of 1cp vs 2cp overflow.

Now for probabilities with Ruthlessness, no RvS, plus Shadow blades. In this case the only way to get to 5 combo points is with 2 non-proc sinister strikes, the chance of that is 64%. The remaining 36% have overflow. In this case your probability of 1cp overflow is 32% and 2cp overflow is 4%.

For Ruthlessness + RvS, the probability of overflow is 20%. In all overflow cases, the number of combo point overflow is 1.

For no RvS, no ruthlessness, the chance of overflow with 3 Sinsters Strikes is 100%  For one combo point overflow, chance is 51.2%, 2cp overflow is 12.8%. For 2 Sinister Strikes chance of no overflow is 32%, chance of overflow of 1 cp is 4%. Sanity Check: 51.2+12.8+32+4 = 100.

For no Ruthlessness, with RvS+SB chance of overflow, probabilities come down to how many SS it takes to get to 3 combo points. This means 20% of the time no overflow (SSx1 with proc). The remaining 80% of the time there is overflow of 1cp 80% of the time and 2 cp 20% of the time. This means no overflow 20%, 1cp overflow 64%, 2cp overflow 16%. 20+64+16 = 100.

Now to put it all together, we need to know frequency of SnD refreshes (for reasons of ruthlessness) and shadow blades uptime. Looking at a typical simcraft run with shuriken toss (as a baseline) we have a shadow blades uptime of 29.48%, Eviscerate count & interval of 60.6 @ 6.51 sec, Slice and Dice of 16.8 @ 24.42 sec, Rupture of 14.6 @ 26.01 sec, and Revealing Strike interval of 24.07 seconds. Five-point finisher frequency is at least 1 / (1/6.51 + 1/26.01) = 5.21 seconds per finisher (SnD excluded because only rupture & evis are guaranteed 5-point finishers). Assuming a priori that SnD has an equal probability to be any number of combo points, the percentage SnD that grants an extra cp is that 60% of every evis/rupture that follows a SnD does have the ruthlessness cp. The Slice and Dice count is 16.8, so this means that 6.72 of the total damaging finishers do not start with ruthlessness, or 91% do. Damaging finishers per minute is 11.5, RvS per minute is 2.4. .

I’m not going to show the whole string of math, but it comes down to this. The total wasted combo points per minute saved by Anticipation, analytically computed (using some empirical statistics from simcraft–specifically Shadow Blades uptime and finisher/RvS frequency) is 0.236 combo points per damaging finisher, and 0.223 wasted combo points per Slice and Dice. By fight length this is a total wasted combo point tally of 21.4, or 3.2 per minute.

Fortunately, Simulationcraft can track wasted combo points. This particular sim wasted 23.4 combo points, in an average fight length of 401.36 seconds, or 3.5 combo points per minute. Not quite in agreement, but pretty close.

Ok, so Anticipation is saving us on the order of 3 combo points per minute, shown empircally through statistic and analytically through modeling. Marked for Death generates 5 combo points per minute (but also wastes approximately 1 per minute because of ruthlessness), so the net gain with Marked for Death is 1 combo point per minute. Seems pretty close. So I decided to run some sims with Shuriken Toss as a control, Anticipation, and Marked for Death casting it <= 1 combo point, through 5 combo points to see at what point Marked for Death is not worth casting.

Here are the simulation results:
Shuriken: 490824
Anticipation: 498824
MfD1: 505516
MfD2: 505463
MfD3: 504375
MfD4: 503229
MfD5: 500878

Yes, I was as suprised as you probably are right now reading that. MfD used on cooldown is a gain over anticipation regardless of your current combo point status. How can this be, if it is wasting combo points. Well, the answer to this is simple. Not every time it is cast is there 2-5 combo points. The looser your combo point requirements, the more frequenty your Marked for Death usage. That said the frequency increases by less than the amount of wasted combo points. This lead me to revise my intuition regarding whether the benefit of Marked for Death truly is the combo points, so I delved into simulationcraft a little more depth.

With Marked for Death we are generating free combo points. Combo points generally cost energy and energy is a limited resource. This means by adding combo points for free, we are shifting some of our limited resource (energy) from combo point builders into combo point finishers. Combo point finishers deal much more damage per energy than builders, so the end result is that our energy spenders are doing more damage on average.

Armed with this logic, I generated a spreadsheet to demonstrate this by seeing exactly what the contribution to energy is. You can view this spreadsheet here. If you look at the spreadsheet, it is clear to see that Marked for Death is incredibly wasteful in terms of combo points but it is shifting energy expenditure from Sinister Strike into Eviscerate, which has double the damage per energy. By granting free combo points and shifting energy consumption into finishers we are also increasing our energy availability through relentless strikes. This is evidenced by looking at the RPS spent values. Also note the huge jump in combo points consumed by talenting Marked for Death. This means finisher frequency is significantly higher. This is because those combo points are free.

In conclusion, Marked for Death’s advantage is not combo point efficiency, but about energy efficiency.