The everyday blog of Richard Bartle.

RSS feeds: v0.91; v1.0 (RDF); v2.0; Atom.

Previous entry. Next entry.

7:34pm on Wednesday, 12th February, 2014:

Ninja Metrics


I bumped into Dmitri Williams at the conference yesterday, and he spent an hour showing me what he's been working on at his ninjametrics company.

Essentially, he has a system that combines statistics and AI machine learning techniques to calculate how much of an individual player's life-time value is due to gameplay and how much it's due to social factors (that is, if I buy something and you later buy it, and this happens often enough, it will realise that I have an influence on you so some of what you spend is because of me, not because of the gameplay). For some people, the social factor is low input but high output: these are guild leaders and so on, with lots of contracts who respect their opinions. Others may have no social component but spend a lot: these would be the lonely whales. Others have a negative social component: these are the trolls and griefers who (in large numbers) might drive people away.

This is interesting enough, but it gets better. The system can also tell how likely someone is to leave in whatever time period you specify, and how much you'll lose if they do. Someone might have another $40 in them if they didn't quit but have an 80% chance of leaving; they could be worth courting before they do. Someone with another $1.50 of lifetime value might be worth an email, but you're not going to send them a box of chocolates through the post to try win them round. This is the kind of information that games companies (and indeed any ecommerce company) should be extremely keen to obtain. Its predictions have been proven correct in actual games, too: if you look at all the people with an 85% chance of leaving within 21 days, then 21 days later 85% of them will have left.

I'm not a fan of using metrics to drive game design. However, using metrics to help customer service is another matter entirely. You could use the NinjaMetrics system to make game mechanic changes (for example by giving guild leaders a superior loot table for lockboxes), but that's not its intended use. You could also apply it to other situations. I did ask if it could be used for law enforcement, but found that actually the relationship had been the other way round: from applying the system to a database of the drug trade, NinjaMetrics got the data they needed to develop a system to find gold farmers in MMOs..!

No, I've not been paid to write this and I don't have shares in NinjaMetrics. I just want better games. If your definition of "better" is "more fun", then this is a systems that can do that without ever touching the mechanics. That's different, and this is what makes is special.

Latest entries.

Archived entries.

About this blog.

Copyright © 2014 Richard Bartle (richard@mud.co.uk).