Please upgrade your browser for the best possible experience.

Chrome Firefox Internet Explorer
×

Balanced Ideal Tank Stats

STAR WARS: The Old Republic > English > Classes > Roles
Balanced Ideal Tank Stats

KeyboardNinja's Avatar


KeyboardNinja
04.14.2015 , 07:19 PM | #41
Quote: Originally Posted by dipstik View Post
i would think you would get 6 points. defend/resist, kinetic weapon shielded, kinetic force shielded, kinetic weapon unshielded, kinetic force unshielded and internal.
Kinetic Weapon and Kinetic Force shielded and unshielded are the same values, respectively. Different probabilities, but the same values. Since I'm grouping by squish, those values disappear.

Incidentally, noob question here: should I just be summing these probabilities when I group? I'm pretty sure they're definitionally disjoint.

Quote: Originally Posted by dipstik View Post
i like that first curve a lot. that looks like a guassian curve
Mathematica's interpolation gets all the credit there. :-)

Unfortunately, due to grouping and interpolating, my curves are dramatically less interesting than yours. (insert innuendo of choice)

Edit You may find this interesting: http://imgur.com/xTJcfIF.png :-) It's strange, because I would think that this graph would be a lot more useful pivoted to the side, with mitigation up-down and probability left-right. However, that particular data pivot produces an incredibly dense graph full of lots of ups and downs and no clear conclusion. This graph is deceptive, but seems to match the gut test on how each tank feels. Good call on the axes, dipstik.
Computer Programmer. Theory Crafter. Dilettante on The Ebon Hawk.
Tam (shadow tank) Tov-ren (commando healer) Aveo (retired sentinel) Nimri (ruffian scoundrel)
Averith (marksman sniper) Alish (lightning sorcerer) Aresham (vengeance jugg) Effek (pyro pt)

December 13, 2011 to January 30, 2017

dipstik's Avatar


dipstik
04.15.2015 , 12:12 AM | #42
I have 16 disjoint cases that sum to one, kw_named, kw_bland, kf and ef each coupled with defended, resisted, shielded and full hit.

It looks like you are summing all the probabilities associated with squish=0. The same should should be true for all of like squish.

Also, thanks for the kudos, but you were the first one to graph the probability density graph. Funny, all this time I've been reporting spike, but never thought to graph it.

ps: heres the numbers for all the bosses assuming 6k dps, k=2, using squish to optimize unless 5*dps*(squish (with relic) +k* sigma )>0.5*health, in which case i optimize squish+k*sigma.

Code:
sin		Malfar	Squad	Undrlrk	Cmndrs	Retrnd	Sparky 	Bulo 	Torque 	Mstr	Cortani
	d	185	185	997	605	185	185	823	582	185	1374
	s	1928	1908	1574	1676	1928	1901	1558	1922	1908	1928
	a	1728	1910	1432	1722	1728	1917	1622	1499	1910	539
	base	0.4394	0.2641	0.2234	0.2825	0.4004	0.3471	0.2764	0.3391	0.3463	0.4513
	D	0.4354	0.2603	0.2188	0.2780	0.3972	0.3431	0.2720	0.3349	0.3434	0.4497
	A	0.4372	0.2591	0.2186	0.2785	0.3970	0.3436	0.2725	0.3354	0.3422	0.4455
	S	0.4373	0.2589	0.2188	0.2784	0.3971	0.3435	0.2724	0.3358	0.3421	0.4473
	relic	WD	WS	WA	WD	WA	WD	WD	WD	WS	WA
	=ward	16944	7621	8402	9545	11593	10616	9759	10296	9298	9556
	sigma	0.2848	0.2115	0.2062	0.2446	0.2697	0.2634	0.2461	0.2698	0.2531	0.2576
	b-mods	9.0	0.0	0.0	0.0	9.0	0.0	0.0	0.0	0.0	9.0
	optimize	1.0051	0.2589	0.2186	0.2780	0.9364	0.3431	0.2720	0.3349	0.3421	0.9608
											
jugg		Malfar	Squad	Undrlrk	Cmndrs	Retrnd	Sparky 	Bulo 	Torque 	Mstr	Cortani
	d	356	356	1644	1564	356	1388	1869	356	417	1548
	s	1928	1838	1200	1200	1928	1200	1200	1765	1915	1928
	a	1557	1809	1159	1239	1557	1322	934	1781	1671	365
	base	0.4519	0.2620	0.2262	0.2852	0.4048	0.3510	0.2799	0.3526	0.3457	0.4459
	D	0.4478	0.2581	0.2223	0.2819	0.4014	0.3482	0.2768	0.3476	0.3427	0.4443
	A	0.4500	0.2584	0.2227	0.2822	0.4021	0.3483	0.2766	0.3501	0.3425	0.4415
	S	0.4501	0.2582	0.2226	0.2820	0.4022	0.3480	0.2767	0.3499	0.3425	0.4429
	relic	WD	WD	WD	WD	WD	WS	WA	WD	WA	WA
	=ward	20306	9963	10704	11897	14298	14030	11998	14648	12054	12918
	sigma	0.2748	0.1891	0.2054	0.2426	0.2556	0.2582	0.2451	0.2597	0.2339	0.2456
	b-mods	9.0	0.0	0.0	0.0	9.0	5.2	0.0	5.6	0.0	9.0
	optimize	0.9974	0.2581	0.2223	0.2819	0.9126	0.3480	0.2766	0.3476	0.3425	0.9326
											
											
pt		Malfar	Squad	Undrlrk	Cmndrs	Retrnd	Sparky 	Bulo 	Torque 	Mstr	Cortani
	d	185	185	920	568	1913	415	953	1304	185	1913
	s	1928	1928	1730	1912	1928	1928	1713	1928	1928	1928
	a	1728	1890	1353	1523	0	1515	1337	609	1864	0
	base	0.4648	0.2318	0.2105	0.2755	0.4194	0.3461	0.2721	0.3495	0.3303	0.4579
	D	0.4614	0.2285	0.2065	0.2716	0.4179	0.3429	0.2685	0.3467	0.3277	0.4568
	A	0.4629	0.2281	0.2065	0.2720	0.4138	0.3428	0.2686	0.3452	0.3272	0.4525
	S	0.4627	0.2274	0.2063	0.2718	0.4156	0.3427	0.2684	0.3461	0.3266	0.4542
	relic	WD	WS	WS	WD	WA	WS	WS	WA	WS	WA
	=ward	18327	10363	9558	10339	10123	11917	10822	11131	12520	10447
	sigma	0.2981	0.1904	0.1794	0.2381	0.2710	0.2690	0.2398	0.2743	0.2583	0.2644
	b-mods	9.0	0.0	0.0	0.0	9.0	8.0	0.0	9.0	1.4	9.0
	optimize	1.0576	0.2274	0.2063	0.2716	0.9558	0.3427	0.2684	0.8938	0.3266	0.9813
the numbers for 4500 dps:

Code:
sin		Malfar	Squad	Undrlrk	Cmndrs	Retrnd	Sparky 	Bulo 	Torque 	Mstr	Cortani
d		1567	185	997	605	185	185	823	1142	185	185
s		1200	1908	1574	1676	1901	1901	1558	1496	1908	1908
a		1236	1910	1432	1722	1917	1917	1622	1366	1910	1910
base		0.4313	0.2641	0.2234	0.2825	0.3977	0.3471	0.2764	0.3376	0.3463	0.4328
D		0.4289	0.2603	0.2188	0.2780	0.3945	0.3431	0.2720	0.3343	0.3434	0.4305
A		0.4292	0.2591	0.2186	0.2785	0.3945	0.3436	0.2725	0.3341	0.3422	0.4298
S		0.4292	0.2589	0.2188	0.2784	0.3945	0.3435	0.2724	0.3343	0.3421	0.4297
relic		WD	WS	WA	WD	WD	WD	WD	WA	WS	WS
=ward		18436	7621	8402	9545	12000	10616	9759	11406	9298	12642
sigma		0.2962	0.2115	0.2062	0.2446	0.2720	0.2634	0.2461	0.2753	0.2531	0.2686
b-mods		0.0000	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000
optimize		0.4289	0.2589	0.2186	0.2780	0.3945	0.3431	0.2720	0.3341	0.3421	0.4297
											
jugg		Malfar	Squad	Undrlrk	Cmndrs	Retrnd	Sparky 	Bulo 	Torque 	Mstr	Cortani
d		2005	356	1644	1564	1152	1353	1663	1707	421	356
s		1200	1838	1200	1200	1421	1314	1200	1200	1810	1838
a		798	1809	1159	1239	1430	1335	1140	1096	1772	1809
base		0.4397	0.2620	0.2262	0.2852	0.4008	0.3496	0.2799	0.3423	0.3456	0.4350
D		0.4376	0.2581	0.2223	0.2819	0.3984	0.3469	0.2765	0.3394	0.3426	0.4326
A		0.4378	0.2584	0.2227	0.2822	0.3984	0.3469	0.2769	0.3397	0.3426	0.4328
S		0.4379	0.2582	0.2226	0.2820	0.3983	0.3467	0.2767	0.3396	0.3424	0.4327
relic		WD	WD	WD	WD	WS	WS	WD	WD	WS	WD
=ward		20209	9963	10704	11897	15910	13938	12145	14472	12770	16498
sigma		0.2900	0.1891	0.2054	0.2426	0.2626	0.2580	0.2443	0.2711	0.2350	0.2520
b-mods		0.0000	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000
optimize		0.4376	0.2581	0.2223	0.2819	0.3983	0.3467	0.2765	0.3394	0.3424	0.4326
											
											
pt		Malfar	Squad	Undrlrk	Cmndrs	Retrnd	Sparky 	Bulo 	Torque 	Mstr	Cortani
d		1352	185	920	568	371	470	953	1032	185	185
s		1495	1928	1730	1912	1928	1928	1713	1671	1928	1928
a		1157	1890	1353	1523	1704	1605	1337	1300	1890	1890
base		0.4588	0.2318	0.2105	0.2755	0.4009	0.3437	0.2721	0.3439	0.3299	0.4362
D		0.4567	0.2285	0.2065	0.2716	0.3983	0.3407	0.2685	0.3409	0.3273	0.4342
A		0.4569	0.2281	0.2065	0.2720	0.3983	0.3407	0.2686	0.3410	0.3268	0.4340
S		0.4567	0.2274	0.2063	0.2718	0.3980	0.3404	0.2684	0.3408	0.3263	0.4336
relic		WD	WS	WS	WD	WS	WS	WS	WS	WS	WS
=ward		18350	10363	9558	10339	14909	12523	10822	12835	12688	17149
sigma		0.3050	0.1904	0.1794	0.2381	0.2849	0.2707	0.2398	0.2793	0.2587	0.2849
b-mods		0.0000	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000
optimize		0.4567	0.2274	0.2063	0.2716	0.3980	0.3404	0.2684	0.3408	0.3263	0.4336

KeyboardNinja's Avatar


KeyboardNinja
04.15.2015 , 12:02 PM | #43
Here is the full set of mitigation cases, plotted over all mitigation profiles: http://imgur.com/jGlkh92.png

I think I actually have a pretty good intuition for why this graph represents spikiness now. Notice how Vanguards (and to a lesser extent Guardians) basically "sit' in the higher probabilities. Their mitigation may not be quite as good overall, but their cases are all in nearly the same "probability stratum", and that stratum is quite elevated. Thus, Vanguards will, with roughly equal probability, mitigate varying percentages of hits. Because it is roughly equal probability, a healer (or anyone else looking at the health bar over time) is just going to "get used" to seeing the fluctuations in mitigation. Nothing unusual happens, because nothing is unusual! Since everything is nearly the same probability, nothing is out of the ordinary.

Shadows are the opposite. They have some really awesome mitigation cases, and it's easy to see looking at the graph how their mitigation can average out to be higher than a Vanguard. However, their probabilities are all over the map. There are some low probability cases which look very different from the high probability cases. Contrary to what the intuitive reading might imply, the two points in the center are not a good example of this (since the casual observer wouldn't ever see a difference between these points). A better example is the cluster pair of points on the far left compared to the apex of the curve in the middle. The apex is high probability and quite good mitigation. Off to the right is almost as much of a contrast, with consistently one of the highest probability max-mitigation cases (Guardians are nearly even, but slightly ahead).

Anyway, this is basically the same intuition that we can derive for why shadows have the lowest standard deviation. In both cases, the "sharpness" of the graph is more interesting than the precise magnitudes, and I think that both graphs (but especially the cases one) do a pretty good job of reflecting player intuition on how the tanks feel.
Computer Programmer. Theory Crafter. Dilettante on The Ebon Hawk.
Tam (shadow tank) Tov-ren (commando healer) Aveo (retired sentinel) Nimri (ruffian scoundrel)
Averith (marksman sniper) Alish (lightning sorcerer) Aresham (vengeance jugg) Effek (pyro pt)

December 13, 2011 to January 30, 2017

dipstik's Avatar


dipstik
04.15.2015 , 12:36 PM | #44
i agree. i think the spikiness of a tank pertains to the probability variance, and psychology thereof, more than the squish variance.

gearing to reduce sigma does serve to reduce the variance in probabilities it seems. this should give healers a better health flux of the tank to respond to, even though they are taking more damage.