It all depends on the context of what you are doing but:
Designing it around flat values is viable if you don't wish the competitive factor to be heavy. = Win or lose vs anyone gets you the same result, no need to seek out better players to play againts.
FLAT AND PERCENTILE:
You can take percentages of existing points but then you have to work on engagement rules. Is a 0 hour noob going to be able to take a shot at someone with a lot of points. Games that open engagements fully mean that an observant but sneaky party can wait for stronger people to spend their forces fighting each other, to only sweep easy targets. No right or wrong - just choose if you want that kind of social dynamic or not. Either way you have to give flat points because 122142% of 0 is still 0.
You can track all the points people ever had, or the largest amount of points people ever had. Then draw upon that source only or parallel to any other points you deal.
This way defeating prestigious players still nets a result even if they are now low on current points. Limiting the exploit-ability of afk/ish players is as simple as only draw points from record IF the provider of points has more than 0 current.
In any event. A tool is complimentary to a goal. If you have a specific idea of what you want to achieve you will be much better suited to chasing down details. Even is said idea is as abstract as "I want the player to get points very hard, but lose them easily - so the ones with much points can feel way superior".
Otherwise you fall pray to the design loop of "we ought to have it because we think its probably popular/good to have it"... Instead of because we want to make something of it. Use it to fill a ... Make it feel like ...
Who are you going to buy? "I don't know, we saw that making TVs is trendy so we opened a factory" or ... "We felt we wanted to improve This and That, to make TVs better, so we went into the business..."