How the routes are made

1. World model. Every quest the faction can do, with its level, prerequisites, givers and enders, and the work its objectives take: kills at the mob's spawn clusters, item drops at their recorded rates, objects, escorts. Dungeon, raid, PvP and elite quests are filtered out, so a route is something one player can do alone.
2. Search. For each start (Orc, Tauren, Undead and Skyborne for the Horde; Human, Dwarf, Night Elf and Skyborne for the Alliance), randomised greedy construction, local search and simulated annealing evaluate millions of complete 1-60 routes, restarting until the best route stops improving or a cap on evaluated routes is reached.
3. Route rules. When routes are compared, grinding counts 2.5 times and every move into another zone counts 7.5 minutes, so a route takes a hub's quests, does them in one loop and moves on rather than chasing a few experience points across the map. Reported hours stay real.
4. Consensus. Each start keeps its own route to level 12 (the starting zone guides); from there one shared route per faction is searched, so the guides join into one chain.
5. Guides. The consensus route is exported as skeleton guides, one per zone visit, and the guides are written from them by hand. They drop side trips for one or two small quests behind a flight, boat, hearth or long ride; where that leaves the level short, a step names a camp to grind at (with an optional dungeon for the long ones). The finished guides are scored under the same model, next to the best route found.

The numbers#

Hours to 60, solo, under the model. The routes were searched four times, independently: the first search is the one the guides were written from; each later one used another random seed, more routes and newer data. Each search is scored on its own world. The guides are scored on the world of the data this release ships with (5eef72140b61), beside those of the first release.

The guides are scored on what they require. Their optional extras (quests started by rare drops such as the OOX beacons, the Green Hills of Stranglethorn pages, profession and bag quests, rares) are left out, because most players will not get them; the last column counts them as done, which is how these hours were published before 0.2.4 (issue #30 (opens in a new tab)).

Read the rest of this section571 words

Start	Search 1: best / consensus	Search 2	Search 3	Search 4	Guides 0.2.4 (grinding)	Guides 0.1.0 (grinding)	Guides 0.2.4 with every optional extra
Orc	119.4 / 120.5	120.7 / 117.8	117.7 / 116.8	119.6 / 117.3	131.1 (34.2)	128.5 (29.7)	122.8
Tauren	120.4 / 120.2	119.1 / 117.4	118.3 / 116.5	117.1 / 116.8	130.1 (34.2)	127.8 (29.1)	121.9
Undead	120.8 / 120.6	119.0 / 118.1	119.3 / 117.0	118.8 / 117.4	131.5 (33.8)	128.9 (29.1)	123.3
Skyborne (Horde)	120.2 / 120.1	117.3 / 117.3	116.3 / 116.3	116.8 / 116.8	130.1 (32.8)	128.3 (28.2)	122.3
Human	119.6 / 119.1	120.7 / 118.8	119.2 / 116.8	119.3 / 118.9	132.5 (37.0)	132.1 (35.5)	127.5
Dwarf	119.5 / 119.5	120.8 / 118.8	117.4 / 117.3	119.8 / 118.8	133.1 (36.9)	132.6 (35.5)	128.2
Night Elf	120.0 / 119.5	120.9 / 118.9	119.0 / 117.5	119.1 / 119.1	133.0 (37.1)	132.6 (35.6)	128.1
Skyborne (Alliance)	121.2 / 119.6	118.6 / 118.6	118.2 / 117.3	119.0 / 118.4	133.5 (36.2)	132.5 (34.5)	129.0

So a Horde character following the guides should expect about 130-132 h, an Alliance one about 132-134 h; doing the optional extras as they come can save up to about 8 h (Horde) and 5 h (Alliance). With every optional extra counted, the guides of 0.2.4 and 0.1.0 are within 0.6 h (Horde) and the Alliance guides of 0.2.4 are 0.8-1.6 h faster; counting only what each requires, 0.1.0 comes out 1.8-2.6 h faster for the Horde because the later guides marked more of their quests optional.

The charts show only the guides: the searches' best and consensus routes were searched on earlier worlds, and a chart draws only numbers from one world.

How close to the best? Nobody can prove a route is the best one possible, so the claim is narrower and checked:

The searches land in the same place

Four independent searches (3.5 to 30 million complete routes per start) find routes within 116.3-121.2 h to 60, and their consensus routes within 116.3-120.6 h. The top of the search space is flat: many different routes finish within about 2% of each other, and most searches were still finding small improvements when they reached their cap on routes, so later searches keep shaving minutes, not hours.

The winners hold up when the model is wrong

Re-scored under 2000 random draws of the model's parameters, the best route found stays within 1% of each draw's best in 76-100% of draws (fourth search), and no route wins every draw: "best found" names one of several near-equal routes.

The guides trade hours for a smooth, safe run

For what they require they take 13.3-14.1 h (Horde) and 13.6-15.1 h (Alliance) longer than the fourth search's consensus (scored on its own world). Almost all of that is grinding the guides choose: side trips for one or two small quests behind a flight, boat or hearth are left out, and quests that need a group, a rare drop or a level the route has not reached are optional or gone, so the guides grind about 33-34 h (Horde) and 36-37 h (Alliance) where the searched routes grind under 1 h and about 8 h. The Alliance mid-40s have few solo quests left once those trips are gone.

Dungeons do not beat questing solo under the model: searched with dungeons allowed (Orc, Tauren, Undead), no dungeon route came out ahead.

The whole method, every parameter and its known limitations are in docs/SIMULATION.md.