Investgame analysis
Skill InvestGameLtd/investgame-skills/skills/investgame-analysis
Use whenever a request needs more than a plain data pull - anything analytical about gaming companies, deals, or investors: valuation and precedent multiples, deal benchmarking, comps and peer sets, premiums, trends and dynamics over time, historical activity, "how active / how much / how does this compare". Triggers: "what's X worth", "value this studio/deal", "comps for a mobile studio", "precedent multiples", "EV/Revenue for [segment]", "valuation benchmark", "what multiple did similar deals get", "is this deal expensive or cheap", "how has [segment] deal activity trended", "interpret these financials", "read this round". It guides HOW to approach the task: scope it, pull the right numbers, read them, and present a rounded, well-footnoted answer. Works hand-in-hand with investgame-gaming-data (the hub, for the data) and investgame-format (for presentation). Not for game-design or generic software questions.From its SKILL.md
npx -y skills add InvestGameLtd/investgame-skills --skill investgame-analysisAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
12.0 KB, as published. Nobody here has run it
InvestGame Analysis
The analytical brain for hard questions. The investgame-gaming-data hub tells you WHAT exists and how to reach it; Analysis tells you HOW to handle a request that needs more than a plain pull - comps, multiples, valuation, premiums, trends, dynamics over time. Your job is to scope the question, pull the right numbers, read them like an analyst, and hand back a rounded, transparent answer.
First, classify the request
Decide what KIND of question it is before you do anything:
- Consolidated (aggregate) - a count, total, average, ranking or trend across many deals or companies ("how many companies above 10x", "average revenue multiple for mobile M&A", "deal value by year"). The database answers these DIRECTLY - it counts, sums, averages, groups and ranks across the whole dataset in one go, including over valuation multiples - so don't pull a long list and tally it by hand; ask for the aggregate itself.
- Record-level - specific named deals or companies and their details ("the multiples on the X acquisition", "Y's funding history"). Either way, think in two layers: which records are in scope, and what to derive from them. The thing YOU own is the scope: a consolidated question is usually unanswerable until you pin it down - "how many companies above 10x" needs which companies (segment, platform, type), which multiple, and which period first. Settle the scope, then ask for the number. (One practical limit: a very wide breakdown - many countries, years or genres at once - can be truncated; a single count or average is always exact. For a wide breakdown, narrow the grouping or say it's a top-N.)
Clarify before a big analysis - then commit
When a request is broad or could be read several ways, don't guess: state your reading in a line and offer a couple of concrete options ("pure-mobile studios, or anything with a mobile presence?"). Once it's clear, execute - don't keep asking. Be helpful, not annoying.
- "comps" / "peers" means companies of the same TYPE and same SEGMENT as the target - ideally the same platform, or at least the same B2B/B2C side. If someone just says "comps", confirm the peer definition before running a comprehensive set.
Multiples - say what you mean, and never break silently
- Pin the metric. "Multiples" can mean EV/Revenue, EV/EBITDA, EV/EBIT or EV/Cash-EBITDA. People say "revenue" or "EBITDA" loosely - confirm or state which you used, and always show WHICH multiple each number is. Pin the EV basis too: a multiple is EV-based and "EV" means the Upfront Enterprise Value (100%-basis at announcement), never the Max / transaction EV (earn-outs excluded); a private round uses post-money EV, a listing uses market cap. State the basis you used.
- The never-break-silently rule. Gaming deals usually disclose only ONE profit metric. If the one asked for isn't there, show the next available one WITH A NOTE - don't return nothing. For profit multiples, prefer EBITDA, then EBIT, then Cash-EBITDA (by how often it's disclosed) unless the user asked for a specific one. For periods, prefer LTM (trailing twelve months), then CY0 (the calendar year OF THE ANNOUNCEMENT, not the current year, written with a zero), then NTM (forward). Example note: "Mostly EV/EBITDA LTM; for a few deals only EV/EBIT was disclosed, flagged in the table." Silently dropping a deal because the exact metric is missing is worse than showing the nearest one with a note.
- Same idea for connected fields. If the exact field is blank but a related one is present, show the related one and note it - e.g. if deal size is undisclosed but the enterprise value is there, show the EV (and vice versa). Show the relevant neighbour, not something unrelated, and keep it balanced.
Read it like an analyst
- Pull comprehensively, then filter. Any single multiple/period is sparse, so a one-metric request collapses to a misleading handful. Ask for the whole multiple set across periods at once, then narrow.
- Summarise honestly. Report median AND mean, the range, and how many deals are in the sample - and say which you led with (a single mega-deal drags the mean). "NM" (not meaningful) and blank (undisclosed) are both left out of the statistics but kept in the displayed list; neither is zero.
- Decide the helpful supporting fields. Part of your job - NOT the formatting skill's - is choosing which EXTRA facts around the answer help the user decide: alongside a multiple, the deal date, acquirer, target segment, deal size; alongside a fundraising, the round, lead investor, stage. Add what's analytically useful, stay relevant, don't dump unrelated data. (HOW it's laid out is investgame-format's job.)
- If there's nothing to compare, say so. If the peer set is empty or too thin, never invent a number. Widen one dimension at a time and label it ("no pure-VR deals, so this is broader gaming"), fall back to the nearest segment as a flagged proxy, or state there's no reliable comparison and stop.
Precision rules that change the answer
-
Multiples are measured on the deal's enterprise value, not its headline size. For an acquisition that's the Upfront EV (never the Max Deal Value, which includes earn-outs); for a private round it's the post-money value; for a listing it's the market cap at listing. The headline "size" means different things across deal types - don't use it as the multiple base.
-
"mobile-only" is not "mobile". Pure-mobile means the platform is mobile only; "mobile" (inclusive) includes cross-platform companies, which are priced differently. Confirm which the user means, or run both and show the gap. This single distinction changes the entire comp set, so resolve it explicitly:
Ask Peer set Why it matters "mobile-only" / "pure mobile" platformis exactlyMOBILE- single-platform, no PC/consolethe clean, like-for-like set; safest for a mobile-native target "mobile" (inclusive) MOBILEis present inplatform(may also includePC_CONSOLEetc.)broader; mixes in cross-platform companies that trade on different economics Conflating the two corrupts the median. Default to confirming; when in doubt, run both and show the gap.
-
Count distinct deals - a deal with several investors must not be double-counted in a market-wide total (a per-investor tally legitimately credits the full deal to each participant). When a sum and a count appear together, restrict both to disclosed-size deals so the two reconcile.
-
Don't blend deal types in one figure without saying so. Control acquisitions, minority stakes, rounds and listings answer different questions; if the set is mixed, say so and split where it matters.
-
A multiple is a ratio - comparable across currencies, never "converted". Absolute figures (EV, deal value) are reported in US dollars; label any that isn't.
Live public-market context - a companion, never the source
When you benchmark against a listed company, you can enrich the read with its current public-market context
(live share price, market cap, today's trading multiples) via investgame-public-markets (the public-market
skill; it owns the live-market tool). Use it to add context, not to replace the proprietary read:
- Lead with the InvestGame deal/company finding; public-market figures only enhance it and must never override or contradict an InvestGame number.
- They measure different things - a historical deal's Upfront EV vs a listed peer's market cap today - so state which is which rather than comparing them as if they were the same.
- It is InvestGame market data. Never name, describe or speculate about where the public-market figures come from; the source stays invisible to the user.
Be objective, not opinionated - but fair
Most of the time the user wants the information; give it cleanly and let it speak. When they explicitly ask you to analyse, or for your read, go deeper and offer a fair, evidence-based view of what the numbers show (a premium or discount to peers, and why), without overclaiming - frame it as what the data indicates, not a verdict.
Always run a critical-thinking pass before you answer
Before returning anything, re-check it from a different angle - devil's advocate on your own work. Did I scope it the way the user meant? Did I pick the right peer set, metric and period? Are there deals I wrongly included or excluded? Is the sample thin enough that I should caveat it? Have I footnoted every assumption and scope choice? Think harder, then answer.
Footnote your work
Whenever you make an assumption or a scope choice, state it: which platform definition you used, which deal types, which metric and period, median vs mean, any fallback you applied. A reader can only trust and compare an answer when they can see how it was built.
Output
Lead with the read (where it lands and why), then a supporting table with the helpful fields, then the summary stats and a one-line note on scope and method. Use investgame-format to present it on-brand.
Reading a multiple (plain-English)
A multiple is how many times a company's revenue or profit the price represents - "8x revenue" means the price was eight years of sales. Higher means the market expects more growth or quality; lower means slower growth, more risk, or a weaker asset. What lifts a gaming multiple: fast growth, strong margins, a hit franchise, recurring live-service revenue, scarcity. What lowers it: single-hit risk, declining bookings, platform dependence, a distressed sale. EV/Revenue is the most-disclosed and the usual default; profit multiples matter more for mature, profitable studios. Always read a number against the peer median, never on its own.
Reading a trend (over time)
Never read deal count alone - read count, total value and valuation (multiples) together: count can rise while value falls (more, smaller deals), and a value spike is often one mega-deal, not a real shift. So normalise: call out any single deal that dominates a year's value, and remember disclosed value understates the truth because many deals report no size - a rising count with flat value may just be lower disclosure. Always caveat the most recent period: a part-year or not-yet-settled quarter looks like a fall purely because it is incomplete. State the window and whether the latest point is partial.
Common mistakes
- Asking for one multiple -> thin or empty result. Pull the full set, then filter.
- A mean with no median, range or sample size - one mega-deal skews the mean.
- Dropping a deal silently because the exact metric/period is missing - show the nearest with a note.
- Conflating "mobile-only" with "mobile" - different universes; confirm, or show both.
- Valuing off the headline deal size instead of the enterprise value.
- Treating "NM" or blank as zero - leave both out of stats, keep them in the list.
- Inventing a number when there are no comparable deals - say "not enough comparables" instead.
- Letting live public-market context override the InvestGame read, or naming where it comes from.
- Doing the analysis but not footnoting the assumptions and scope - the reader can't trust what they can't see.