Guides

@Deluxe That's not stubborn at all and I buy bonus fodder all the time!

Consider a bonus fodder at =1250t above the floor to be equal to a non-bonus fodder at the floor (as long as you didn't forget to exalt on time). Subtract the bonus from the cost to be the equivalent "floor price" - so, an 8500t bonus dragon acts as if it cost 7250t without bonus. Then, compare these times.

With the current estimates, it looks like level 7 is good for fodder costing 7250t! (If you really really don't like level 7s, level 6 is sort of close at that price point.) Though, level 7s with the current times are slightly faster in speed to level 6s on a non-50%-EXP-boost date. (My Drakeharvest level 7 time is 1 minute 14 seconds, while my non-EXP-boost level 6 avg. time is 1 minute 18 seconds. Keep in mind my Drakeharvest trials have far less volume, so the average hasn't yet stabilized.)

At a non-bonus cost of 7250, you would be making ~3093 profit per minute for lv6, or 3256 profit per minute for lv7. :)

Enter our second micro-holiday, Frigidfin Expedition! Like Drakeharvest, we've go a 50% experience boost; this 2018 event lasts Dec 13 06:00 - Dec 17 06:00. Coincidentally, Dec 16, the final full day of this event, will coincide with the first day of a conquest battle between Fire and Shadow. As such, I was curious about mercenary profit margins and performed some speed tests for comparison. Here is the [url=https://docs.google.com/spreadsheets/d/1w7S1ag5Ff652iisMnbScLbqq_lGZNbsOtKrZ2MqU6gA/edit?usp=sharing]link to the raw data sheet[/url]. * The values below are from a snapshot taken at the time of this post; I may add more data later, which might slightly alter the average times / profit per minute values, but the decisive level profit breakpoints should be the same. At the moment, it's looking like the non-bonus adult floor will be [b]9000t or higher[/b]. I'll include 7000t and 8000t prices as breakpoints to account for the occasional sniping / rollover hatchlings, but keep in mind that they'll be slightly rarer than the floor price adults. My average times for Frigidfin are slightly slower than my Drakeharvest times due to RNG differences. (In Drakeharvest, the data set was tiny and I didn't get any particularly bad luck back then.) My [b]Frigidfin level 6[/b] average time is currently 12 seconds (25%) higher than my [b]regular level 5[/b] average time. Here is a comparison of my Frigidfin times to my regular non-micro-holiday times in the Mire, using a 119 STR / 72 QCK build: [columns] [b]Level[/b] 4 5 6 7 8 [nextcol][color=white]____[/color] [nextcol][b]Reduced avg. time per dragon (hh:mm:ss)[/b] 0:00:24 0:00:39 0:01:00 0:01:27 0:02:17 [nextcol][color=white]____[/color] [nextcol][b]EXP boost reduction (compared to reg. times)[/b] [color=green]-1 sec[/color] [color=green]-9 sec[/color] [color=green]-18 sec[/color] [color=green]-39 sec[/color] [color=green]-75 sec[/color] [/columns] Based on these times and an [b]assumed adult floor of 9000t[/b], I estimated potential mercenary payouts as follows: [code] Level Payout Lv.4 9500t Lv.5 10500t Lv.6 11500t Lv.7 12500t [/code] With those payouts, the profit per minute (in Treasure) based on my personal average times are as follows: [code] Profit per minute in Treasure Assumed payouts Lv5: 10500t, Lv6: 11500t Cost Lv.4 Lv.5 Lv.6 Lv.7 7000 6145 5404 4534 3776 8000 3687 3860 3526 3090 8500 2458 3088 3022 2746 9000 1229 2316 2519 2403 9500 0 1544 2015 2060 [/code] This shows a breakpoint of fodder costing 8500t where level 6 profit starts to equal or exceed level 5 profit. [b]Note that this is only true for the estimated payouts in the table above.[/b] If the level 5 payout was only 10000t, for example, the profit for a level 6 payout of 11500t begins to catch up to level 5 profit as early as a 7000t adult cost. [code] Profit per minute in Treasure Assumed payouts Lv5: 10000t, Lv6: 11500t Cost Lv.5 Lv.6 7000 4632 4534 8000 3088 3526 8500 2316 3022 9000 1544 2519 9500 772 2015 [/code] Going back to profit calculations based on assumed payouts of 10500t lv5, 11500t lv6: once adults hit 9500t, level 7 profit catches up to level 6 profit based on the assumed payouts. However, with conquesting flights that are ordering mercenaries, it probably makes more sense to go for the level volume and ask for level 6 dragons instead. (The difference in profit per minute is very small at that breakpoint anyway.) Ignoring the absolute price of adults and going by relative margins, we can make the following conclusions. Level 6 is better at the following profit margins: [LIST] [*]+1000 per lv5, +2500 per lv6 [*]+1500 per lv5, +2500 per lv6 [*]+2000 per lv5, +3500 per lv6 [/LIST] And Level 5 starts to become marginally better if the margins are: [LIST] [*]+2000 per lv5, +3000 per lv6 [*]+2500 per lv5, +3500 per lv6 [/LIST] Thanks to the reduction in training time, based on the calculated costs and payouts, it looks like [b]level 6[/b] is the most profitable for sellers and most cost/time efficient for buyers while Frigidfin Expedition is active.

Thanks for the data as usual, I love reading your analysis @/Sylvandr! I wonder how fodder floor will turn out after Fire vs Shadow, them having to start at a 10kt fodder floor is really unfortunate, hopefully they can regain some profit while still being able to DOM!

Nice Job, this helped out a bunch

Update 2021-08-05: This is now an outdated post. Please check this post for the updated version.

---

I finally did a massive revamp of my training teams. I hadn't had the time before to verify the right QCK levels and turn orders, but I recently put in about a month of effort and finally decided on the new stats, then took as many speed tests as I could for the different builds.

One thing I noticed is that Coliseum 2.0 has made many solo trainer builds perform very well. The fix in the turn order, as well as the much-reduced buff cast times, have all contributed to much lower times across the board, and surprisingly, more consistent times. I tested several different builds between Mire and Ghostlight Ruins, and even though I change my playstyle quite a bit depending on the build, I ended up with really similar times for all of them.

Ideally I'd have 100+ trials (200 dragons) per level for levels 4-6, and closer to 200+ trials (400 dragons) per level for levels 7+, because that's about what it takes for the variant RNG times to stabilize into an average that doesn't move when you have a particularly good or particularly bad time. However, I can't commit that much time to collecting (I don't want to train 400 level 25s per build...) so I did the best I could. I'll list my numbers for each build first, then get into a time comparison and some notes.

I'm arbitrarily choosing to list the builds in order of my favorite to least favorite, but you'll see in the time comparisons that they're all quite good builds and you should be able to choose based on preference.

Post with builds and turn order KO threshold comparisons

Ghostlight Ruins Meditate high QCK - 115 STR/64 QCK/40 VIT/10 DEF Water
Speed test sheet

Level 4 5 6 7 8

____

Time per dragon 21 sec 44 sec 1 min 11 sec 1 min 57 sec 2 min 57 sec

____

# trials 79 (158 dragons) 78 (156 dragons) 76 (152 dragons) 28 (56 dragons) 25 (50 dragons)

Mire Scratch - 119 STR/63 QCK/31 VIT/8 DEF Arcane
Speed test sheet

Level 4 5 6 7 8

____

Time per dragon 24 sec 44 sec 1 min 13 sec 1 min 55 sec 2 min 54 sec

____

# trials 122 (244 dragons) 124 (248 dragons) 88 (176 dragons) 41 (82 dragons) 30 (60 dragons)

Ghostlight Ruins Meditate low QCK - 115 STR/53 QCK/48 VIT/21 DEF Water
Speed test sheet

Level 4 5 6 7 8

____

Time per dragon 21 sec 44 sec 1 min 13 sec 1 min 57 sec 3 min 2 sec

____

# trials 107 (214 dragons) 106 (212 dragons) 55 (110 dragons) 14 (28 dragons) 1 (2 dragons)

Ghostlight Ruins Scratch high QCK - 115 STR/65 QCK/39 VIT/8 DEF Water
Speed test sheet

Level 4 5 6 7 8

____

Time per dragon 20 sec 45 sec 1 min 12 sec - -

____

# trials 26 (52 dragons) 27 (54 dragons) 26 (52 dragons) - -

And now I've got a few notes for the build comparisons.

1) "Ghostlight Ruins Scratch high QCK" has the least amount of trials in this list. Its "20 second" lv4 average is not necessarily 20 seconds because it hasn't had time to stabilize; the other two GLR builds started around 20 seconds as well, but eventually the effects of bad RNG runs evened out the times to 21 sec each. I expect the Scratch version to level out towards 21 sec as there are more trials.

2) With any of these builds, it is reasonably easy to kill bosses as long as you start with enough Breath and haven't lost too much health. However, I chose not to fight bosses at all when taking these speed tests, regardless of conditions. (Ie: I would flee even if I had full health and full breath, which could easily get me a boss kill.) This is because I did not want additional randomness to be thrown into my speed tests; you can't control whether you're in the condition to kill a boss whenever it appears, and there's a possibility of a random miss or crit causing you to lose when you should've won. This really skews average times. In addition, assuming that you weren't worried about average times, even if you only fought bosses under good conditions, they take so many turns to kill that their "large" exp reward is slower than just running into some 3-packs. Also, they need more than a full Breath bar to kill, so unless you're really lucky with crits, you'll be starting your next fight with not enough Breath for Eliminate. This adds time to the average even under ideal conditions, and I did not find that useful for speed test comparisons. (Outside of speed tests, I fight bosses if I'm not in a hurry, but I'll still avoid them if I'm trying to train quickly.)

3) I experienced a much higher incidence of bosses in the Ghostlight Ruins compared to the Mire. This observation is supported by data in the enemy frequency sheet. The averages on the sheet show the GLR bosses at 2.7% and 2.1% appearance rate, while the Mire bosses are at 1.4% and 1.2%. What I noticed is that this frequency difference is enough such that GLR "bad RNG runs" are caused by a mix of missed Eliminates AND bosses appearing, while Mire "bad RNG runs" are mainly limited to poorly-timed missed Eliminates. Although bosses did sometimes appear in the Mire, their presence was not as devastating to my average times as the GLR bosses.

4) Just like it was in Coli 1.0, Ghostlight Ruins (any viable build) is significantly better than Mire for level 4s due to the pack configurations.

5) I did not test GLR Anticipate for 2.0 because the turn order has become rather hostile to the Anticipate style of play from my point of view. Dropping QCK will help reduce the number of wasted turns, but you don't want to drop too many turns so that you're still fine in the non-startup rounds. I also didn't test Anticipate or Meditate builds in Mire because I feel Scratch would have a slight edge due to the pack configuration, but I can't say for sure unless I test them. Ideally, I would test these other builds eventually. However, I left them low on my priority list due to the amount of time and effort required.

Here are the times side-by-side. Keep in mind that a lower number of trials means that the average might move a few seconds in the long run.

Build GLR Med high QCK Mire Scratch GLR Med low QCK GLR Scratch high QCK

______

Lv.4 21 s 24 s 21 s 20 s *

___

Lv.5 44 s 44 s 44 s 45 s *

___

Lv.6 1 m 11 s 1 m 13 s 1 m 13 s 1 m 12 s *

___

Lv.7 1 m 57 s 1 m 55 s 1 m 57 s * -

___

Lv.8 2 m 57 s 2 m 54 s - -

* Asterisk denotes low number of trials relative to the other builds.

My conclusion: The builds are so close together that it's up to you whether you go to Mire or Ghostlight Ruins. Scratch and Meditate seem both good for GLR. You could also go for either more QCK or more VIT/DEF. (I gave specific numbers for the highest defensive stat combo available with remaining points, which allows you to take the most number of hits.)

For my personal choices, I'll be doing lv4 - 6 with GLR Meditate high QCK. For levels 7 and up, I'll use Mire Scratch.

Hope this helps a bit with build choices! We have many more choices thanks to 2.0.

@Sylvandyr

Thank you for testing these builds! I prefer the Scratch versions and it's good to know the battle times are similar.

@Maki I was surprised how similar the times became! I'm glad to know that people have lots of options out there. :)

My guess is that the versions I didn't test would probably still be in the ballpark, though I did pursue the ones that I thought had the potential to be the "fastest" first. One day I'll collect tests for Anticipate GLR/Mire and Mire Meditate, and fill in the remaining holes. :P

Nice! I keep coming back to meditate monk as my favorite build, glad to see it's up there as one of the fastest. I just realized though that I've been using one with even higher qck: 76. Maybe I'll do some trials with that and with these two variations to see how it stacks up.

@Thoracosaurus 76 is the old 1.0 QCK setting. I did this overhaul on my own Coli teams to drop unnecessary QCK! I listed the useful KO thresholds in one of my links.

@Sylvandyr

Nice, good to know! Thanks!