Tonight's national address by President George Bush marks the start of the endgame in Iraq. By this time next year, we will almost certainly know whether the U.S. has won or lost its gamble on a new stable democratic Iraq. We will certainly know this on the morning after the next U.S. presidential election that is now only 22 months away.

Not that fighting will have ended in Iraq by either date. Insurgency and counter-insurgency operations may then be even more intense. In particular, if U.S. troops are withdrawing, both Sunni and Shia militias may perhaps harry the retreating forces. They will certainly devote more effort and ruthlessness to killing each other in the battle to control post-American Iraq.

When the British left India in 1947, they did so in good order and without fighting. The estimated million Indian casualties that followed were slaughtered by other Indians in ethno-religious pogroms by both Hindus and Muslims. The same will happen in Iraq if the U.S. withdraws in defeat.

It is already happening on a smaller scale. U.S. Democrats argue that the present Sunni-Shia fighting is caused by the U.S. presence. Withdrawal, they argue, would compel all Iraqis to settle their differences by compromise.

This argument is so spectacularly at odds with the facts -- principally the fact that the U.S. presence restrains rather than provokes Sunni-Shia warfare -- that it has to be an argument made in bad faith.

As Rich Lowry of the National Review points out, Democrats want to withdraw but they are unwilling to make honest arguments in favour of withdrawal. So they take refuge in manifestly insincere claims such as U.S. provocation and go on to suggest "there is no military solution." The truth is that there can be no political solution until someone establishes a clear military predominance in Iraq generally and Baghdad in particular. Neither Sunni nor Shia politicians will reach compromises if they are certain to be murdered by the militias of extremists on their own side. Peace and stability require both a military and a political solution -- but the political rests on the military one rather than vice versa.

That said, there are only three politico-military forces that could plausibly win the kind of military victory that would determine the political shape of Iraq. The first is a "national unity" Iraqi government supported by U.S. and other coalition troops. In other words, the status quo -- though one that can and should be marginally improved by ditching some extremists to make the government and its armed forces more genuinely national. This force remains by far the strongest one in Iraq and, if the U.S. remains in the mix, it could not conceivably be defeated by either of the other two forces. Mr. Bush's speech tonight is expected to outline a new approach that aims at actually defeating the insurgents: some combination of more troops, more money for reconstruction and job programs, a new U.S. military leadership, and a different set of tactics to "hold" troubled areas after they have been "pacified."

These policies are reasonable. They build on the current position of strategic dominance -- troubled, certainly, but dominant nonetheless. And they have a good chance of success, but only if the Bush administration makes clear that they are not a temporary surge designed to accelerate an early withdrawal but represent a firm U.S. commitment to establish a friendly stable government in Baghdad. Without that commitment, the terrorists will fight on until the deadline, official or imagined, even if they are losing badly.

On the other hand, unless the coalition obtains some early success against the insurgents, domestic American pressure for a U.S. withdrawal will eventually become overwhelming. And once Americans start withdrawing, the Iraqi "national" army will disintegrate into ethnic factions, and a battle would begin between the two other forces -- the Shia and Sunni militias.

In these circumstances, the second force, the Shia militias, would be the heavy favourites to emerge victorious from the brutal civil war following withdrawal. They are more numerous -- the Shiites are 60 per cent of Iraq's population. They would have the backing of an increasingly Shia-dominated government and thus favourable access to heavy weapons. And they are already in possession of weapons and money from Iran.

Other things being equal, the Shia militias would defeat their opponents within a year or two although guerrilla warfare would continue, perhaps indefinitely, in mainly Sunni districts. Because the Shia would enjoy their victory through militias rather than through the ballot box, they might end up with an outright Shia dictatorship rather than with a democracy weighted towards Shia interests as now. At any rate, such a dictatorship is more likely as the outcome of an outright Sunni-Shia civil war than of the present mixture of democratic manoeuvring and urban violence restrained by American troops.

The third force, the Sunni militias, have no realistic prospect of victory. If they were being massacred following withdrawal, however, some of the neighbouring Sunni Arab regimes -- Jordan, Egypt, Saudi Arabia -- might feel compelled by their own public opinions to intervene militarily. That would, at best, secure a small Sunni enclave while inviting all sorts of unpredictable disasters -- possibly an Iranian intervention, a Kurdish declaration of independence, a Turkish incursion, etc.

As Reuel Marc Gerecht establishes in an authoritative and sobering article in the current Weekly Standard, however, these and other disasters going a great deal wider geographically would be invited by the mere withdrawal of U.S. troops before a coalition victory.

The other disasters include massive ethnic cleansing on the Indian model, vast refugee incursions into Jordan and surrounding states and serious threats to friendly pro-Western regimes as a result, the collapse of American power in the region, the acquisition of nuclear weapons by Egypt and the Saudis as a consequence of being deprived of the American nuclear umbrella, a renewed Islamist threat to Muslim regimes as far afield as Pakistan and, finally, a revival of morale in al-Qaeda as battle-hardened terrorists become available for terrorism in the Middle East, Western Europe, and elsewhere.

What are the powerful reasons for risking these dangers? If the Democratic leaders were put on a truth serum and asked why the U.S. should leave Iraq, they would probably reply "because we have lost." If the U.S. and its allies have lost, however, then who has won?

They cannot answer that question. No one has won as yet. No one is actually winning unless it is the U.S. So why surrender?

John O'Sullivan is a senior fellow at the Hudson Institute in Washington and editor-at-large of the National Review.