The Gambling downtime rules in XGtE state that it takes the character a workweek (5 days) and a stake of between 10gp and 1000gp to partake in the gambling downtime activity.

The resolution of this activity (according to the XGtE rules) is three skill challenges are attempted by the player, against three randomly generated DCs (DC = 5+2d10 for each check).

Wizards of the Diviner School have the feature Portent which gives them:

When you finish a long rest, roll two d20s and record the numbers rolled. You can replace any attack roll, saving throw, or ability check made by you or a creature that you can see with one of these foretelling rolls. You must choose to do so before the roll, and you can replace a roll in this way only once per turn.

Each foretelling roll can be used only once. When you finish a long rest, you lose any unused foretelling rolls.

As the gambling downtime takes place over 5 days, the question arises over when exactly do these checks occur? Is it all on a single day, spead out over the 5 days?

This question is important, as it directly affects how much of the Portent ability is available for the player to use.

My instinct is to spread the checks over the 5 days, and use this method:

1. Player declares in advance how much they want to gamble (their stake)


2. The DM rolls a set of DCs for the three checks for each day (using the 5+2d10 method)


3. The diviner would then roll their portent for each day, and choose one or more of the checks (if any) to use their portent on. They can also choose just to chance it and do any number of "regular" rolls on a given day.


4. Passing or failing the check would be resolved by comparing the result of their roll(s) against the generated DCs for the day for the relevant checks.
   
   
   
   
   • If a check is failed or passed on a given day it is passed or failed for the whole session
   
   
   
   


5. If by the end of the 5 days they still have outstanding checks left to complete, these would be resolved by straight ability check rolls against the "average" DC for the session (calculated for each skill using 5 + Average(DC/day - 5)). These checks would not be eligible targets for Portent as they are multi day checks.

Would this ruling be balanced (and/or exploitable)?

To be clear here, the houserule is the daily "sets" of DCs being generated and the "average" set at thee end, instead of the set XGtE makes out for the 5 day activity as a whole.

That ruling heavily favors the diviner

When you're rolling DCs privately and randomly, this has no real effect on the player's choices. Because I have no information on what you've rolled behind the DM screen, it does not affect my actions. As such, we'll ignore the randomness of the DC and just focus on the wizard's side.

Assuming level 2 divination wizard, they will have 2 foretelling rolls per day, totalling 10 rolls over the 5 days. Because you allow them to forgo making any checks on a day they don't like their Portent chances, this is only slightly worse than "roll 10 take highest 3" (it's slightly worse because if you get 15s one day and forgo the check in the hopes of something higher, you can't go back to 15). This kind of selection is very strong.

This is not to say it's unbalanced, because you as the DM set what happens. If you're in a setting where magic is normal, people might know exactly what diviners can do, and even if you aren't, consistently "lucky" hands will tip off anyone that pays attention.

You might represent this as a higher than 10 percent chance to run into a gambling complication, or the overriding of a chance, and you just say "it happens."

However,

When you make the checks isn't clear

So your ruling is as valid as any, because it's left up to DM purview.

The abstractions break here

You are in a situation where the abstraction that is the downtime gambling rules hit against the abstraction that is powers usable only at certain frequency. This is what happens when any associated rule is stretched far enough.

When the abstractions break, the solution is to look at the fiction and represent the effect in a different way.

Here, gambling over a long time period is represented by a series of dice rolls. It is not true that the dice rolls happens at particular points in fictional time; rather, they are a way of representing something on a coarser scale than what the rules otherwise default to.

Look at the fiction

Consider, together with the group, what is a reasonable edge the diviner wizard would have in games that are partially of chance, and how would you go ahead representing that mechanically.

The game has a default solution for when this is the case - the character gets advantage. So, unless you have a particular reason to do something else, just give the character advantage and be done with it.

If making or losing money by gambling turns out to be central to the game, build a more elaborate set of rules for it, and build them so that they handle situations on the scale of days, so that the abstraction of gambling and the abstraction of long rests do not clash. But most games would manage without.

Whether the checks occur all on the same day or on different days is up to DM fiat.

Buuuuut..... It might be pretty strange, narratively speaking, if each check all occurred on the same day, yet the downtime activity still occupied a full week. Strictly RAW, you're allowed to space out the checks however you like, but for the sake of narrative cohesion, it may be appropriate to have these checks occur on different days.

As a result, the Wizard would be able to reroll their Portent Dice each day, and would gain the opportunity to use their portent dice on each of these checks.

As DM, you can allow them to make all three checks on the same day if they get lucky with their portents (and want to risk it on the last one), but if it were me, I'd not allow the player to "defer"/"expedite" those checks; the Portent feature is already a powerful advantage on these checks, and giving the player the freedom to wait until they get especially good Portent rolls is going to make their gambling ability especially powerful.

Of course, a lot of this depends on how important the Gold Economy is in your campaign. If a character being able to nearly guarantee earnings of 1k gold per workweek is not a big deal, then you can do whatever is most convenient for the player without too much issue. It makes narrative sense for a Divination Wizard to be uncannily good at Gambling, so while the ability to earn a lot of gold very quickly for relatively minimal risk is a powerful feature, it does make narrative sense, and if you're not worried about exploitability, then you can structure the rules in whatever form is most convenient for the player.

You can't cheat in downtime gambling.

Downtime assumes you're doing safe, consistent, repeatable activities where the details aren't interesting enough to roleplay out.

What you're doing here is running a scam, and the details are highly relevant: which bookie are you going to clean out? What defenses do they have? How much are you going to try to take them for? And most importantly, do you get caught?

This is a volatile situation and needs to be played out at the table. Frankly, a lot of players will want to play it out at the table, because everyone loves a good casino heist.