One of the three modes of play that these rules support is free-form play, where every contest is resolved with a single opposed roll. For some GMs and groups, this by itself is enough. Many successful campaigns have been run in this way, ever since it was adapted for RPGs from the style of wargame invented by Tom Mouat.

However, sometimes you might want a contest that isn't resolved so quickly. One that fills more of the session, one that has several twists and turns that generate an emergent story without the GM having to work too hard. And sometimes you might just want to roll more dice.

Chaosium's Questworld SRD has 4 distinct mechanisms aimed at filling that gap (group, chained, wagered and scored sequences). These rules propose a 5th, which in Questworlds terms would be called an ongoing sequence, but we call a ongoing contest.

To understand why these rules are the way, we go back to the probability table produced for opposed rolls:

We could use the direct equivalent of a QW scored sequence on that table; repeatedly rolling until 5 successes have been scored by one side or the other. As in QW, we use a tiebreak  rule of 'highest roll wins'.  Without such a rule, the number of rolls required is unbounded, and sometimes excessive. With such a rule in place, every roll moves you at least one step closer to completing the contest. So there can't be more than 10, and are commonly less than 5.

This would lead to:

This table has several problems. For one thing, it has large areas where the chance of the underdog wining is less than 5%. This is because an unexpected result might occur once, but it is less likely to show up repeatedly. 

Consequently, it is sufficiently different from the single-roll table that players may be tempted to try and persuade the GM to use this type of contest when it favors them. Worse, it favors them most when they are dealing with inferior opposition, and penalizes them when they are the underdogs. This is the opposite of the typical use of the two types of contests for pacing and drama. The climax of a session or campaign, the crux to be focused on, would normally the harder conflict. The boss fight, not the easier job of dealing with the guards.

Things become worse once one or two successes have been scored, and the stronger side now has a 99.99% chance of winning. Unless the GM realizes this and cuts things short, the conflict must still proceed for several more rolls in order to ensure the overwhelmingly likely thing does indeed happen.

The solution adopted for ongoing contests is to have two different types of rolls:

• A decisive roll represents a test to definitively resolve a contest, determining a winner, using a tie-break mechanism if required. 
• An task roll represents a test to see if specific attempt as an individual challenge or task is successful. 

A contest is complete not at a fixed threshold, but because one side decided to attempt a decisive roll. Typically, this will be because they think they are in an advantageous position. This approach places the climax, the decisive moment, at the end of the contest. Which is where it belongs.

As this involves active decision-making, you can't really give a definitive single table for the probabilities involved. The table below comes from using a fixed tactic of 'decisive roll if ahead, task roll if behind but not yet lost'.

This has a less steep gradient than the scored contest table, much closer to the original..It has a slight bis to the PC, assuming hey are the ones making the decisions as o what roll to attempt.