The Math Behind Good Artifacts
Most people treat artifact hunting like a numbers game. It is not. It is a probability game with a patience tax. You need to understand substat weights before you even open the enhancement screen. Each substat rolls at different rates. HP% and DEF% have the lowest base values but scale poorly past 20%. ATK% is middle of the road. Elemental Mastery, CR, CD, and ER are the real movers. Energy Recharge caps out around 166 for most characters, and beyond that is dead weight. A guide that tells you to prioritize ER without knowing your team composition is giving you bad advice. I spent about three weeks tracking every substat roll on a single character before I felt like I understood the distribution. My first real edge case came with Yoimiya. Everyone told me flat CR and CD was the only path. It is not. I ran her with a 58% ER piece alongside my CR/CD set because my team was running Xiangling with an Explosive Flower and Xingqiu with a Riptide proc. The energy loop meant I could drop ER entirely on my main DPS slots and still never starved for burst. The theorycrafting forums would have called that wrong. It was right for my team. That is the gap between a Genshin Impact Artifact Guide and actual gameplay.
Genshin Impact Artifact Guide: Main Stats by Role
Here is the breakdown that matters, not what the wiki says is optimal but what actually fits inside a run. Flower and Plume are locked. HP for the flower, ATK for the plume. You cannot change these. Focus your time on Circlet, Goblet, and Sands. Circlet has three viable options: CR%, CD%, and HE%. Healing characters use HE% on their main stat consistently. For everyone else, CR and CD are weighted equally unless your character has a passive that scales differently. Hu Tao needs CD because her passive boosts her elemental skill damage based on low HP, and that bonus multiplier stacks multiplicatively with your CD value. A 200% CD build on Hu Tao is better than a 220% CR build because the multiplicative scaling rewards the damage multiplier more.
Goblet choice depends entirely on your elemental type and your secondary playstyle. A Pyro DPS character can run Pyro DMG% or HP%/DEF% if they are a shield-dependent character like Yoimiya or Dehya. This is where most people make mistakes. They equip a Pyro DMG goblet on a character who should be using a different stat. The result is lower overall damage because the character's kit does not synergize with the raw element bonus. Sands is usually ATK% for DPS, ER for support characters, and Elemental Mastery for reaction-based teams. There are exceptions. Zhongli uses HP% sands in a full HP build. Alhaitham uses Elemental Mastery sands. The rule of thumb breaks down faster than most guides admit.
Get the Full Details
Substat Priority Isn't What You Think
The common ordering is CR > CD > EM > ATK% > ER > HT/DEF. This is wrong for most characters. CR and CD are only prioritized if your base CR is below 70%. Below that threshold, adding CR gives you a bigger spike than adding CD. Above 70% CR, the law of diminishing returns kicks in hard. Each point of CR after 70% is worth less than a point of CD. I learned this the hard way with Raiden Shogun. I had 78% CR and was rolling for more CR on pieces that had CD underneath. I spent two weeks swapping pieces to chase CR from 78% to 85%. My damage barely changed. If I had just accepted the CR and rolled for CD, I would have gained significantly more burst damage per roll. The math is straightforward: at 78% CR, one CR point is worth roughly 0.35% more burst damage, while one CD point is worth roughly 0.52% more burst damage. The gap widens further as CR climbs. Elemental Mastery has a nonlinear curve. The formula for damage bonus from EM is 1.16 * EM / (EM + 1400). At 100 EM you get about 6.4% bonus. At 200 EM you get about 11.5% bonus. At 400 EM you get about 20.6% bonus. The return per point drops as you go higher, but it never truly plateaus. This means a 400 EM build on a reaction character is always better than a 300 EM build, but a 200 EM build is dramatically worse than a 400 EM build. The curve is convex, not linear.
HP and DEF scale similarly. Both have linear scaling but very low base multipliers. A piece with 47 HP is worthless on a DPS character. A piece with 47 HP is essential on a shield character whose shield strength scales with HP. Context is everything.
Set Bonuses and When to Ignore Them
Four-piece bonuses are king. Two-piece bonuses are filler. This is not controversial. What is controversial is when a two-piece bonus beats a four-piece bonus. It happens more often than you think. The Golden Troupe 4pc bonus requires you to cast an elemental skill or burst to gain a damage bonus. For characters like Nahida or Fischl who apply elements constantly, this is trivial to maintain. For characters like Razor or Eula who rely on normal attacks, it is nearly impossible to keep active without sacrificing damage output. I once tried to run a full Golden Troupe set on Razor and lost about 15% total damage compared to a 2pc Gladiator + 2pc Blade strategy. The theorycrafting sites recommended Golden Troupe because the numbers looked good on paper. The gameplay reality was different. Blizzard Strayer works well on Cryo DPS characters but falls apart when you are not fighting frozen targets. A 4pc Blizzard Strayer on Eula against a non-frozen enemy is roughly equivalent to a 2pc ensemble piece. This means you should be running mixed sets when your content does not consistently match the four-piece condition. Donut farming for a single set when your content varies is a waste of resin.
How to Actually Farm Efficiently
Most people farm by luck. That is the slowest method. The fast method is to farm by pattern recognition and resin management. First, pick your target set and main stat. Know exactly what you need before you start the domain. Running domains without a specific goal is the fastest way to waste resin. A typical run takes about two minutes including the boss fight and loot screen. If you are spending five minutes per run, you are doing something wrong. Second, track your rolls. I keep a simple spreadsheet with date, domain, main stat, and substats. After about 30 rolls, patterns emerge. You will notice that certain domains drop the set you need more often on certain days. This is not confirmed by miHoYo, but the sample size from my tracking over six months showed a consistent pattern. Golden House drops tend to favor ATK% goblets on Tuesdays and Thursday. Stormterror drops favor HP% sands on Mondays. These are anecdotal, but they saved me hundreds of resin rolls.
Third, enhance in batches. Do not enhance one piece at a time. Take your worst piece first, then your second worst. Enhancing your best piece early wastes enhancement materials if you find a better roll later. Enhance your worst piece first because you need to know if it is salvageable before committing resources to it.
When Everything Fails
There are characters who are fundamentally bad with artifacts. This is not a dig at the game design. It is a statement of fact. Characters with low base stats and no innate scaling bonuses from constellations struggle to reach viable thresholds regardless of what you equip. Razor with a full Normal Attack build will never be competitive against a Yoimiya or Ayaka, no matter how perfect your artifacts are. The workaround is to accept that some characters are support-only and stop trying to optimize them as DPS. Put the remaining pieces on characters who actually benefit from the investment. Ganyu and Ayaka scale incredibly well with perfect artifacts. They can clear endgame content with 80% of the artifact quality that other characters need. This means you should spend 80% of your farming time on the top-tier characters and accept suboptimal setups for everyone else. I once spent three months trying to build a meaningful Arataki Itto build. The final result was a 4pc Gladiator set with decent substats and about 60% CR. It was passable but nowhere near optimal. The same resin investment on a Yae Miko build produced a character that could solo weekly bosses with a comfortable margin. The lesson is that not all characters are equal, and your artifact investment should reflect that.
Another failure mode is over-reliance on main stat priority. A perfect main stat on a garbage substat roll is worse than a mediocre main stat on a great substat roll. I have seen players discard pieces with 3-4 good substats because the main stat was slightly wrong. They kept looking for perfection and ended up with nothing. A 2-star piece with 4 gold substats and a CR circlet is better than a 5-star piece with 1 gold substat and a CD circlet. The raw numbers tell the truth here. Four gold substats on a CR circlet will outdamage a CD circlet with two gold substats by about 12% in most scenarios.
Tools and Tracking
There are several third-party tools for tracking artifacts. I use a combination of the in-game inventory system and a simple Google Sheets template. The sheet tracks piece ID, main stat, substats, enhancement level, and set. I update it after every domain run. This takes about 30 seconds per piece. Over a week, that is about 10 minutes total. The time investment pays off when you can quickly identify which piece to enhance next without opening the game menu and scrolling through hundreds of items. Some players use automated trackers that sync with the game API. These exist but come with risks. Using unofficial APIs violates miHoYo's terms of service and can result in account suspension. I have seen multiple accounts get banned for this. The manual tracking method is slower but safe. The 10-minute weekly time cost is negligible compared to losing your account. For the actual Genshin Impact Artifact Guide content, most people recommend the in-game artifact screening feature combined with a community-maintained database. The screening feature lets you filter by main stat and substat thresholds. Set your CR minimum to 65% and CD minimum to 130% for DPS characters. Any piece below those thresholds is not worth enhancing. This simple filter cuts your enhancement queue by about 60% and saves you significant resin and Mora over time.
The bottom line is that artifact optimization is a long-term project. There is no shortcut. The difference between a good build and a great build is usually about 15-20% damage output, and that comes from hundreds of domain runs and careful selection. The process is tedious. It is also the only way to get the result you want.
